1
DEPARTMENT OF HEALTH AND HUMAN
SERVICES
FOOD AND DRUG
ADMINISTRATION
CENTER FOR DRUG EVALUATION AND
RESEARCH
ADVISORY COMMITTEE FOR
PHARMACEUTICAL SCIENCE
Advisors and Consultants Staff
Conference Room
2
PARTICIPANTS
Arthur H. Kibbe, Ph.D., Chair
Hilda F. Scharen, M.S., Executive
Secretary
MEMBERS
Charles Cooney, Ph.D.
Patrick
P. DeLuca, Ph.D.
Meryl
H. Karol, Ph.D.
Melvin
V. Koch, Ph.D.
Marvin
C. Meyer, Ph.D.
Gerald
P. Migliaccio (Industry
Representative)
Cynthia R.D. Selassie, Ph.D.
Nozer Singpurwalla, Ph.D.
Marc Swadener, Ed.D. (Consumer
Representative)
Jurgen Venitz, M.D., Ph.D.
SPECIAL GOVERNMENT
EMPLOYEES
Judy Boehlert, Ph.D.
Paul H. Fackler, Ph.D., (Acting
Industry
Representative)
Thomas P. Layloff, Jr., Ph.D.
FDA
Ajaz Hussain, Ph.D.
Chris
Joneckis, Ph.D.
Robert O'Neill, Ph.D.
Keith
Webber, Ph.D.
Helen Winkle
3
C O N T E N T S
PAGE
Call to Order:
Arthur Kibbe, Ph.D. 4
Conflict of Interest Statement:
Hilda Scharen, M.S. 4
Introduction to Meeting:
Helen Winkle 8
Subcommittee Reports:
Jurgen Venitz, M.D., Ph.D. 34
Parametric Tolerance Interval Test
for Dose Content Uniformity:
Ajaz Hussain, Ph.D. 44
Moving Forward--An Approach for
Resolution:
Robert O'Neill, Ph.D. 46
Committee Discussions and
Recommendations 53
Process Analytic Technology (PAT)--Next
Steps:
Ajaz Hussain, Ph.D. 70
Finalizing PAT Guidance, Training and
Certification:
Chris Watts, Ph.D. 89
Standards Development:
Ali Afnan, Ph.D. 100
Rapid Microbial Methods:
Bryan Riley, Ph.D. 110
Committee Discussions and
Recommendations 135
Open Public Hearing
Leo Lucisano,
GlaxoSmithKline 135
Parrish M. Galliher,
Xcellerex 145
Troy J. Logan, Siemens 174
Robert Mattes, Foss-NIRSystems 185
PAT Applications for Products in the
Office of
Biotechnology Products (OBP): Overview
and Issues:
Keith Webber, Ph.D. 194
Christopher Joneckis,
Ph.D. 213
Charles Cooney, Ph.D. 224
Kevin Koch, Ph.D. 248
Tom Layloff, Ph.D. 271
Committee Discussions and Recommendations 279
4
1 P R O C E E D I N G S
2 Call to Order
3
DR. KIBBE: Ladies and gentlemen,
shall we
4
begin. This is the Advisory Committee for
5
Pharmaceutical Science. Today is April 13th. Those
6 of
you who have not done your taxes, because you
7 are
here working for us and the Federal Government,
8 you
will get exactly no compensation to allow you
9 to
do your taxes late.
10
Hilda.
11 Conflict of Interest
Statement
12
MS. SCHAREN: Good morning. I am going to
13
start reading the Conflict of Interest Statement
14 for
the Advisory Committee for Pharmaceutical
15
Science. I am Hilda Scharen with
the Center for
16
Drugs, FDA. I am the Executive
Secretary for this
17
committee.
18
The following announcement addresses the
19
issue of conflict of interest with respect to this
20
meeting and is made a part of the record to
21
preclude even the appearance of such at this
22
meeting.
23
Based on the agenda, it has been
24
determined that the topics of today's meeting are
25 issues of broad applicability and there are no
5
1
products being approved at this meeting.
Unlike
2
issues before a committee in which a particular
3
product is discussed, issues of broader
4
applicability involve many industrial sponsors and
5
academic institutions.
6
All Special Government Employees have been
7
screened for their financial interests as they may
8
apply to the general topics at hand.
To determine
9 if
any conflict of interest existed, the Agency has
10
reviewed the agenda and all relevant financial
11
interests reported by the meeting participants.
12
The Food and Drug Administration has
13 granted general matter waivers to the Special
14
Government Employees participating in this meeting
15 who
require a waiver under Title 18, United States
16
Code, Section 208.
17
A copy of the waiver statements may be
18
obtained by submitting a written request to the
19
Agency's Freedom of Information Office, Room 12A-30
20 of
the Parklawn Building.
21
Because general topics impact so many
22
entities, it is not prudent to recite all potential
23 conflicts of interest as they apply to each
member
24 and
consultant and guest speaker.
25
FDA acknowledges that there may be
6
1
potential conflicts of interest, but because of the
2
general nature of the discussion before the
3
committee, these potential conflicts are mitigated.
4
With respect to FDA's invited industry
5
representatives, we would like to disclose that
6
Gerald Migliaccio is participating in this meeting
7 as
an industry representative acting on behalf of
8
regulated industry. Mr.
Migliaccio is employed by
9
Pfizer. Dr. Paul Fackler is
participating in this
10
meeting as an acting industry representative. Dr.
11
Fackler is employed by Teva Pharmaceuticals U.S.A.
12
In the event that the discussions involve
13 any
other products or firms not already on the
14
agenda for which FDA participants have a financial
15
interest, the participants' involvement and their
16
exclusion will be noted for the record.
17
With respect to all other participants, we
18 ask
in the interest of fairness that they address
19 any
current or previous financial involvement with
20 any
firm whose product they may wish to comment
21
upon.
22
Thank you.
23
DR. KIBBE: Thank you. I am Art Kibbe. I
24 am
Chairman of the Pharmaceutical Science's
25
Department at Wilkes University.
7
1
We have a tradition of introducing
2
everyone around the table, so, Dr. O'Neill, if you
3
will start.
4
DR. O'NEILL: I am Bob
O'Neill. I am
5
Director of the Office of Biostatistics in CDER.
6
DR. HUSSAIN: Ajaz Hussain, Deputy
7
Director, Office of Pharmaceutical Science.
8
MS. WINKLE: Helen Winkle,
Director,
9
Office of Pharmaceutical Science.
10
DR. VENITZ: Jurgen Venitz,
Clinical
11
Pharmacologist, Virginia Commonwealth University.
12
DR. SELASSIE: Cynthia Selassie,
Professor
13 of
Chemistry, Pomona College.
14
DR. BOEHLERT: Judy Boehlert. I have my
15 own
pharmaceutical consulting business.
16
DR. SWADENER: Marc Swadener,
retired from
17 the
University of Colorado at Boulder.
18
DR. MEYER: Marvin Meyer, Emeritus
19
Professor, University of Tennessee, now a
20
consultant in Boca Raton, Florida.
21
DR. KAROL: Meryl Karol, a
Professor at
22 the
University of Pittsburgh in Environmental and
23
Occupational Health.
24
DR. LAYLOFF: Tom Layloff,
Management
25 Sciences
for Health, a nonprofit, working primarily
8
1 in
Africa on drug quality.
2
DR. KOCH: Mel Koch, Director of
the
3
Center for Process Analytical Chemistry at the
4
University of Washington.
5
DR. COONEY: Charles Cooney,
Department of
6
Chemical Engineering at MIT.
7
DR. DeLUCA: Pat DeLuca, Professor
of
8
Pharmacy at the University of Kentucky.
9
MR. MIGLIACCIO: Gerry Migliaccio,
Vice
10
President of Global Quality Operations for Pfizer.
11
DR. FACKLER: Paul Fackler, Teva
12
Pharmaceuticals.
13
DR. KIBBE: Thank you.
14
Next on our agenda is an Introduction to
15 the
Meeting. Ms. Winkle.
16 Introduction to Meeting
17
MS. WINKLE: Thank you and good
morning to
18 all
the committee members. I especially want
to
19
welcome the members who have not attended before or
20 are
just joining us for the first time.
21
That includes Dr. Cooney, Dr. Koch, and
22 Dr.
Singpurwalla, who is not here yet, but will be
23
joining us later today, and also to Gerry and Paul
24 for
helping us out as industry reps. We are
really
25
pleased to have both of them here working with us.
9
1
[Slide.]
2
Today, I just want to give a short
3
update--it will probably be longer than short, but
4 it
is supposed to be short--update on some of the
5
things that we are doing in OPS.
6
[Slide.]
7
Today, I want to talk a little bit about
8 the
OPS mission, vision, and goals. I think
it is
9
really important for me to go over these with the
10
committee because it helps all of us understand a
11
little bit more about where OPS is going in the
12
future. I think that as we talk
about various
13 scientific issues, it will help put things in
a
14
better perspective for the committee.
15
We have just recently finalized the
16
mission, vision, and goals, so I think it is
17
important that I share them.
18
I also want to talk a little bit about
19
what we are doing in OPS in developing a new
20
paradigm for CMC review in our Office of New Drug
21
Chemistry. This is really an
exciting effort that
22 we
have undergone, and I think there are a lot of
23
things that will be very beneficial to talk about a
24
little bit here.
25
A lot of this is built on the
10
1
Pharmaceutical Quality Initiative for the 21st
2
Century, so it helps put that in perspective, as
3
well as to what we are doing in the future in OPS.
4
I also want to mention some of the new
5
personnel that we have in OPS and then talk a few
6
minutes about the meeting agenda.
7
[Slide.]
8
The mission statement. Again, I
think
9
this is very important because it sets forth what
10 OPS
is currently focused on, and it is important
11 not
only to those activities that we are engaged in
12 and
working on very diligently in the organization,
13
they are also very important in supporting the
14
overall mission of the Center and mission of the
15
Agency.
16
Basically, our mission statement is to
17 ensure timely availability of high quality
drug
18
products to U.S. patients. We are
doing this
19
through effective and efficient scientific
20
assessment of relevant pharmaceutical and
21
biotechnology information in the submissions, and
22 by
facilitating those scientific and technological
23
innovation that improve understanding of product
24
performance, quality, and efficiency of
25
development, manufacturing, and quality assurance
11
1
processes.
2
Many of these things that we have talked
3
about at past meetings, that we will talk about in
4 the
future, fall very much within this mission
5
statement and some of the things that we are trying
6 to
accomplish.
7
[Slide.]
8
Our vision is to be an international
9
champion. I think it is very
important that we
10
talk about where OPS is going from an international
11
perspective because things are more global.
12
Obviously, now industry, many of the
13
things that we work on are global, and we need to
14 be
part of that overall global involvement in
15
pharmaceutical science, but we really want to be
16
champions and leaders in the regulatory application
17 of
contemporary scientific knowledge, and that
18
knowledge that affects the design, development,
19
manufacture, and clinical performance of
20
pharmaceutical and biotechnology products.
21
[Slide.]
22
Basically, the goals are for OPS programs
23 and
projects to support the achievement of the
24
following attributes of drug products:
25
The drug quality and performance is
12
1
achieved and assured through design of effective
2 and
efficient development and manufacturing
3
processes;
4
That regulatory specifications are based
5 on
a mechanistic understanding of how product and
6
process factors impact product performance;
7
And that there is continuous "real time"
8
assurance of quality.
9
These are all very important objectives
10
that we are striving toward.
11
[Slide.]
12
Also, OPS will implement a review quality
13
system and procedures throughout the organization
14
that will:
15
Recognize the level of scientific
16
knowledge supporting product applications, process
17
validation, and process capability;
18
Apply a risk-based regulatory scrutiny
19
that will relate to the level of scientific
20
understanding of how formulation and manufacturing
21
process factors affect product performance, and the
22
capability of process control strategies to prevent
23 or
mitigate risk of poor product performance.
24
[Slide.]
25
I wanted to talk a few minutes now that I
13
1
have talked about sort of the mission and the
2
goals, and you have a feel for where we are going,
3 I
want to talk about some of the changes that we
4 are
making. Specifically, I want to talk
about the
5
changes we are making in CMC review.
6
To help set the stage for the future, I
7
wanted to go quickly through the FDA Strategic
8
Action Plan that Dr. McClellan initiated when he
9
came on board, I want to talk about the
10
Pharmaceutical Quality for the 21st century, which
11 is
a really important initiative that is taking
12
place in the Agency, and is very important to us as
13 we
move ahead in the Office of Pharmaceutical
14
Science and some of the things that we are trying
15 to
accomplish.
16
I want to talk just a second about
17
resources in our CMC area, because I think without
18
mentioning the resources and the problems that we
19
have in resources, it is hard to understand why the
20
changes are necessary that need to be made in order
21 to
improve on how we do review.
22
Also, I want to talk about a few other
23
influences that have happened since the
24 organization
was first established in 1995.
25
[Slide.]
14
1
The FDA Strategic Plan - Responding to
2
Challenges and Opportunities.
Again, as I said,
3 Dr.
McClellan introduced this plan several months
4
after he entered the Agency. He
was very focused
5
while he was here at the Agency on accomplishing
6
these particular aspects of all of the products
7
that are regulated by FDA.
8
Mainly, he focused on efficient risk
9
management, so that we were sure we were going to
10 get
the most public health bang for our regulatory
11
buck.
12
He wanted empowering consumers.
He felt
13
that I think all of us understand there is a lot of
14
interest on the part of consumers in their own
15
health care, and he wanted to be able to improve
16
health through better information to consumers, so
17 as
they make decisions, as they look at their own
18
health care, as they even deal with their
19
physicians, et cetera, that they have a better
20
understanding of the medications, food, et cetera,
21 et
cetera, that they need to take or use.
22
He wanted to improve patient and consumer
23
safety, protect America from terrorism, and more
24
effective regulation through a stronger workforce.
25
So, as we make changes in OPS, and we look
15
1
toward the future of things that we want to do
2
differently, and how we want to do those, we are
3
trying to incorporate many of the things that Dr.
4
McClellan incorporated in his strategic plan.
5
[Slide.]
6
Also, as I mentioned, the FDA Initiative
7 on
Pharmaceutical Quality is an important
8
groundwork for some of the things that we are doing
9 now
and in the future in OPS.
10
This particular chart is very helpful
11
because it shows the particular dimensions of the
12
plan for strong public health protection, for
13
international cooperation, for risk-based
14
orientation, science-based policy and standards,
15 and
integrated quality systems orientation.
These
16 are
the really important aspects of the initiative
17 and
where we are going.
18
[Slide.]
19
There are various directional vectors that
20
came with the initiative, and I won't go through
21
each of these. I think you can
look through them,
22 but
I think they are important as we look at OPS
23 and
where we are going for OPS in the future, so
24
looking at our regulatory policies, making sure
25
that we incorporate new technology advances when we
16
1 do
our regulation, that we are able to work with
2
industry, et cetera, in doing some of these, and
3
that we have consistency and coordination
4
throughout the whole drug quality regulatory
5
program.
6
[Slide.]
7
Here is basically the directional vectors
8 and
many of the things that are being worked on
9
under the GMP initiative agencywide.
10 These include looking at a
preapproval
11
inspection compliance program, dispute resolution
12
processes being established, a pharmaceutical
13
inspectorate that focuses specifically on
14
pharmaceutical products during the inspection
15
process that is being set up. We
are hoping to
16
have product specialists on inspection process, and
17 we
hope to start that very soon.
18
We have set guidance on CFR Part 11,
19
aseptic processing guidance, a comparability
20
protocol guidance. We have been
doing a lot of
21
stuff with risk management and quality by design,
22
and, of course, PAT, which we have talked about.
23 But
you can see where each of these sits on the
24
whole vector between risk and science.
These are
25 all
important aspects of the initiative.
17
1
[Slide.]
2
But the most important thing to me about
3 the
initiative is it afforded us in OPS, a lot of
4
opportunities to change the way that we do
5
business. It has opened up a
window of time for us
6 to
really look at how we do business and make the
7
changes that are necessary to move forward into the
8
21st century.
9
This is not easy, and I will go through
10
some of the challenges what we have had, but first
11 of
all, I want to talk about some of these
12
opportunities and just mention them to you, because
13 I
think they are really important.
14
We really have the opportunity now to
15
strategize more on how we are going to ensure
16
product quality. This is ensuring
product quality
17
across all of the Center, and it is the first time
18 we
really have thought about the whole aspect of
19
product quality and what needs to be done to ensure
20
in the future that we are focused
on the right
21
aspects of that.
22
We need to revisit our processes.
This is
23 a
really good opportunity for us to do that.
We
24
have built processes over the last 20, 25 years,
25 not
only in review, but in inspection, as well, and
18
1
this gives us an opportunity to look at all of the
2
processes that fall under pharmaceutical quality
3
regulation, and to incorporate best practices.
4
We need to focus more on manufacturing and
5
associated issues relating to the quality of
6
products, one of the things that was very apparent
7 to
us when we went in and looked at the review
8
processes, that we did not pay as much attention to
9 the
actual manufacturing of products and how it
10
affected the quality of the products.
11
So, this is a really good opportunity for
12 us
to do that. We have a lot to learn and
we have
13 to
work with a lot of people because obviously, we
14
don't have as much understanding as we need, but we
15 are
doing a lot and looking at manufacturing
16
science and trying to get a better understanding of
17
that. I think that has been very
apparent in some
18 of
the things that you have talked about with PAT,
19 we
will talk about even more today.
20
We need to focus both on review and
21
inspection, and we need to put more science into
22
those. A lot of times, and it has
been said time
23 and
time again, we have not used really good
24
science in making the decision, and sometimes we
25
have had a lot of complaints from industry and
19
1
others about that lack of really scientific
2
understanding on inspections, so this is a really
3
good opportunity for us to ensure that that science
4
exists, but it is really important that we ensure
5
that it is part of the review process, as well, and
6 it
is going to take time, but I think working with
7 our
people and others, we will get there in the
8
future.
9
We need to enhance the interactions
10
between review, inspection, and compliance. One of
11 the
things that was very interesting to me right
12
before we started the initiative is we met with a
13
number of people from trade associations, and it
14 was
made very clear, the gap between what happens
15 in
review and what happens in inspection, and who
16 is
sitting in the middle but industry with a lot of
17
questions on how policy was set or what the policy
18
means, and dealing with the inspectors day to day
19 who
really don't have an understanding of that
20
either, so we really need to ensure better
21
interaction between review and inspection.
22
We need to foster communication with
23
industry. In the review, we have
been very
24
hesitant to talk much to the industry and to work
25
with the industry, not only on specific
20
1
applications, but on science in general, and there
2 is
a lot of science in the industry that can be
3
very beneficial to us in the Agency to understand
4 the
processes and understand manufacturing and
5
pharmaceutical quality, and we need to do more of
6
that.
7
We need to have early discussion on CMC
8
questions. As I already mentioned, we have a
9
dispute resolution process that we are setting up,
10
which we feel will be very helpful to give industry
11 an
opportunity to talk with us when they have
12
scientific issues or questions.
13
We need to leverage resources for the best
14
bang for the buck. This is a real
problem, and as
15 I
said, I am going to talk a little bit more about
16
resources.
17
We need to simplify the regulatory
18
requirements and we need to be able to find ways to
19
reduce some of the regulatory burden.
We have
20
talked here before at the committee about the
21
number of supplements that we get in the
22
organization, we are really drowning in
23
applications in supplements, and all of them are
24
treated basically the same, and we need to really
25 step
back and look at ways that we can put more
21
1
emphasis or more responsibility on industry and try
2 and
work with them to have better understanding of
3
things, and not get as many applications.
4
We need to eliminate the "check box"
5
approach that we have. What we do
basically in
6
review is we go through and do you have this, do
7 you
have that, do you have this without a real
8
understanding of what the process is, the
9
manufacturing, the whole important aspects of
10
pharmaceutical quality.
11
[Slide.]
12
We need to enhance training opportunities,
13 and
we now have this opportunity under the GMP
14
initiative, as well as some of the things that we
15 are
undertaking in OPS. We are in the
process of
16
working with several of the pharmaceutical
17
industries to set up plant residency programs for
18
some of our chemists.
19
We have other cross-training opportunities
20
that we are discussing, and then we have the
21
pharmaceutical inspectorate, and the reason I put
22
this here is not only will we be able to train our
23
inspectors better as far as some of the aspects or
24
manufacturing science, will it be able to take
25
advantage of those from the review standpoint, as
22
1
well, and I think this will be extremely helpful
2 and
useful to us in our future regulatory
3
activities.
4
[Slide.]
5
We need to enhance FDA's knowledge
6
regarding new technologies in manufacturing, and we
7
need to encourage innovation, and again this goes
8
back to PAT.
9
We need to develop processes that are
10
focused more on product risk, which we have not
11
done. As I said before, almost
every product has
12 the
same weight, same level of review, and we
13
really need to look more at the risk aspects of the
14
product.
15
We need to revisit how quality of products
16
relate to ensuring safety and efficacy, and
17
especially ensuring clinical relevance.
18
We need to alleviate industry's concern
19
regarding reprisal. I hate to put
this up, it's a
20 bad
word "reprisal," but that thought is out there
21
often in industry, I hear it time and time again,
22 and
I am hoping through better interactions with
23
industry, with better understanding of the science
24 and
the ability to discuss the science, we can
25
begin to eliminate some of these concerns.
23
1
We need to enhance our
international
2
involvement. We are working on pharmaceutical
3
development and risk management in international,
4 but
we need to do more of this, because again it's
5 a
very global world out there, and we need to be
6
sure that we are involved in everything that is
7
happening on the international front.
8
[Slide.]
9
I did say I wanted to mention resources
10
real quickly. I thought this
would give you a
11
better perspective again as to why we want to make
12
some of the changes in CMC. The workload is really
13
difficult for our CMC reviewers in new drugs.
14
We got, in 2003, 159 NDAs, 342 commercial
15
INDs, 507 research INDs, 1,858 CMC supplements, and
16
that doesn't include efficacy or labeling
17
supplements, and 1,132 annual reports.
This is a
18 lot
of work to take on, and this is a lot of work
19
because we have fewer and fewer review staff.
20
We have constantly been over the last few
21
years hit by reductions in resources, so we are
22
doing more work with less people, and we have
23
really got to think of ways to streamline the
24
process and to be able to get some of this done in
25 a
more efficient and effective manner.
24
1
[Slide.]
2
Other influences, though, too, that bring
3
about the necessity for change, as I said, in 1995,
4
when ONDC was established, it was collocated with
5 the
clinical divisions, and this seemed to work
6
really well for a couple of years, but a lot of
7
things have happened within the Center, within the
8
Agency, within the world, that really affect how we
9 do
the CMC reviews, so we really need to rethink,
10
based on these influences and changes, how we do
11
things.
12
Some of the influences includes shorter
13
PDUFA deadlines, FDAMA, again harmonization and
14
globalization, such changes in our regulatory
15
processes, such as SUPAC, BACPAC, new technologies
16 in
pharmaceutical manufacturing.
17
[Slide.]
18
PAT, counterterrorism, counterfeit
19 products. We were just talking about the fact that
20 we
can't even begin to keep up with counterfeiting,
21 we
have to find better ways to do that.
22
BSE and other crisis, such as that.
There
23 has
been a greater focus on generic drugs, and
24
tomorrow we will spend a lot of time talking about
25
some of the issues that we have with regulating
25
1
generic products, and it is really important that
2 we
begin to focus more on some of these issues and
3 how
we need to ensure that we incorporate other
4
thinking from the new drug side into how we are
5
going to regulate generic products in the future.
6
There have been a lot of changes in
7
industry, more globalization mergers, et cetera.
8
There has been electronic submissions.
We are
9
working very hard to hopefully enhance the
10
efficiency of our processes through electronic
11
submissions, and there has been more focus on risk
12
management and quality systems.
13
[Slide.]
14
So, basically, what we need to do is to
15
change the paradigm for CMC review.
I have talked
16
about that we have the opportunity to do this. The
17
things that we really need to focus on based on
18
those opportunities is really to develop a
19
risk-based CMC review.
20
I think this is really important, and I
21
think we are going to need help.
This is not going
22 to
be an easy thing to determine risk.
23
I think products are going to come and go
24
that are risky, we see that all the time, products
25
that you don't expect when it comes on the market
26
1 to
have any risk, then, things are found out later
2 on,
so it is not going to be an easy process to
3
develop, and it is going to take a lot of thought
4 and
probably a lot of help even from the committee,
5 but
this is definitely a direction that we need to
6
head in.
7
We need to establish quality systems which
8
help set the framework for ensuring that we do have
9 a
dynamic organization and that we can handle the
10
complications of the regulatory processes.
11
We need to focus resources towards efforts
12
that improve quality, and not hinder and interfere
13
with innovation, and I think that is very
14
important, and we need to focus on all aspects of
15
CMC.
16
We need to look at chemistry, we need to
17
look at manufacturing, and we need to look at
18
controls, and we have not done as good a job of
19
this in the past.
20
[Slide.]
21
The advantages of the new paradigm, for
22
FDA, we will have more product and process
23
knowledge, which can be shared by industry, so that
24 we
have a better understanding of the products that
25 we
regulate.
27
1
We will have more efficient resource
2
allocation for review and inspection, and we can
3
increase our trust and understanding
of industry
4
decision making.
5
[Slide.]
6
The advantages for industry is hopefully,
7
that we will have fewer, more efficient,
8
science-based inspections, faster, more consistent
9
reviews.
10
There is a potential for reduced
11
regulatory burden, for managing changes with less
12 FDA
oversight, for focused resources on critical
13
issues, flexibility to focus on what should be
14
done, not what can be done, and to improve
15
communication with FDA.
16 [Slide.]
17
But most of all, the ultimate beneficiary
18 is
the public, and we hope through some of the
19
changes that we make, that we can increase the
20
availability of drugs on the market, we can have
21
faster approval of new products, we can have
22
continued assurance of high quality products, and
23 we
can increase the public's confidence in the work
24
that we are doing in FDA, and hopefully, reduce
25
costs, which isn't, of course, our business, but
28
1
something we hope is going to come out of some of
2 the
changes that we are making.
3
[Slide.]
4
The new paradigm will include developing
5
strategies to recruit and train reviewers. One of
6 the
things that we realize is that we have a real
7 gap
in the qualifications that our reviewers have.
8
We need more that have understanding of
9
drug discovery, analytical chemistry,
10
pharmaceutical engineering, and we are going to be
11
looking at recruiting and training people in these
12
areas.
13
We need to build a learning organization,
14 one
that is skilled at creating, acquiring, and
15
transferring knowledge. This is
one thing we have
16 not
done an adequate job of in the past, and we
17
really need to work on, probably not only just in
18
OPS, but throughout the whole Center.
19
We need to set specifications based on
20
science and process understanding.
We need to
21
reengineer the process, so that we have the best
22
practices, metrics, and that we are customer
23
oriented.
24
This is another thing that we have not
25
paid a lot of attention to in the past, which we
29
1
really need to look toward in the future, is who
2 our
customers are and what they need.
3
[Slide.]
4
We need to increase emphasis on
5
manufacturing science, we need to ask the right
6
questions at the right time. We
need to implement
7
peer review by FDA scientists and clinicians.
8
Establish a program to better integrate
9
review and inspection, develop processes which
10
ensure regulatory relief based on process
11
understanding and control, quality systems in
12
manufacturing, and continuous improvement is very
13
important, and we need to create a better work
14
environment and promote job satisfaction within our
15
organization.
16
[Slide.]
17
As I said, there is a lot of challenges.
18 The
current culture, both inside and outside of
19 FDA,
is definitely the biggest challenge we have.
20 It
is very difficult to get people to think
21
differently. They have worked in
a certain culture
22 for
years and years, and changing that culture is
23 not
easy. We see that both inside the
Agency, as
24
well as outside.
25
Hiring is not easy, it is very difficult
30
1 to
find people with the right skills that want to
2
come to work for the Government, and this is a big
3
challenge that we have ahead of us.
4
Establishing performance metrics is also a
5
challenge because we have really never had the
6
metrics to measure anything except for the amount
7 of
work we get, and we are really going to have to
8
step back and look at this differently.
9
We need to identify gaps in requirements.
10 We
need to reevaluate the review process again to
11 be
sure we are asking the right questions that
12
ensure product quality.
13
We need to understand what is relevant
14
science.
15
We need to determine what is needed for
16
pharmaceutical development data to assist in a
17
better understanding of manufacturing process.
18
We need to develop a science-based risk
19
model, and we need to integrate
better into the
20
inspection process including
participating on
21
inspections.
22
This is a lot of work we have ahead of us,
23 and
the reason I am sharing it is because I think a
24 lot
of these issues are going to come up in the
25
future where we are going to need the committee's
31
1 input
on how to tackle some of these challenges,
2
some of the things that we need to incorporate into
3 our
review and our processes to make sure that we
4 are
doing what is necessary to have the best
5
regulatory processes available.
6
Again, I feel that this is important that
7 you
all have an understanding of where we are
8
going, and we will look forward to talking about
9
many of these things in the future.
10
[Slide.]
11
Before I go into the agenda, I just wanted
12 to
mention some of OPS's new additions that we
13
have. We are really fortunate to
be acquiring a
14 lot
of new staff lately, and some of the people I
15
think that are very important, that will be working
16
with us very closely, I wanted to talk about today.
17
First, is Dr. Vince Lee. I think
all of
18 you
know Dr. Lee since he was once chair of this
19
committee. We are very happy to
have Vince with
20 us,
and we feel that there is a lot of things that
21 he
is going to be able to help us work on as we
22
move towards changing some of our regulatory
23
paradigms.
24
Also, we will be adding Dr. Mansor Khan
25
from Texas to our staff. He is
going to be our
32
1
director of our Division for Product Quality
2
Research in our Office of Testing and Research, and
3 he
will be joining us next month.
4
We are looking forward, too,
to having Dr.
5
Khan. I think he is going to add
a lot and help us
6 a
lot in some of the areas of research that we need
7 to
be focused on in order to accomplish some of the
8
things that we want to accomplish.
9
Also, I wanted to mention that Dr. Moheb
10
Nasr has become the permanent director of the
11
Office of New Drug Chemistry. I
think many of you
12
know Dr. Chiu has retired. Dr. Nasr so kindly came
13
from St. Louis to take this job, and has been
14
working very diligently on some of the changes that
15 we
are trying to make.
16
Dr. Chi Wan Chen has joined him as the
17
deputy of the office.
18
Also, I wanted to announce that Dr. Keith
19
Webber, who is sitting over here, too, is the
20
Acting Director of the Office of Biotech Products.
21 We
appreciate Dr. Webber stepping in and taking on
22
this very challenging group that has recently
23
joined us in the Office of Pharmaceutical Science.
24
[Slide.]
25
Just to finalize my presentation, I just
33
1
wanted to quickly go through the meeting topics. I
2 think
this is going to be an extremely exciting
3
meeting. I think that the topics
tomorrow are
4
especially stimulating, topics that I think will
5 add
a lot to our future thinking in these areas.
6
Today, we are going to have subcommittee
7
reports. We are going to have a
discussion of the
8
proposal on PTIT. That is
parametric tolerance
9
interval test for dose content uniformity. We have
10
talked about this before. We have
a proposal now
11 on how
we want to finalize our thinking in this
12
area.
13
Then, we want to talk about PAT.
We want
14 to
give an update, talk about some of the things
15
that we have done, and also talk about how PAT is
16
going to be implemented in our Office of Biotech
17
Products.
18
Tomorrow, as I said, I think the topics
19 are
very stimulating, I think we will have some
20
really good discussion on bioequivalence topics.
21 We
want to talk about highly variable drugs, about
22
bioINequivalence. This is very
important.
23
We have a lot of areas here of thought
24
that we need to bring forward and discuss with the
25
committee, and we want to talk about topical
34
1
products.
2
Also, time allowing tomorrow, we have an
3
awareness topic, and this is nanotechnology that we
4
want to introduce.
5
With that, I am going to finish up and
6
hand it over to Dr. Kibbe, and I look forward to
7
hearing the discussion in the next two days.
8
Thank you.
9
DR. KIBBE: Thank you, Helen.
10
We are pretty close to being on time, so
11 we
will turn it over now to the subcommittee
12
reports. The first one is from
Clinical
13
Pharmacology. Jurgen is moving
rapidly to the
14
podium, so here we go.
15 Subcommittee Reports
16
DR. VENITZ: Good morning. I am here to
17
report back from a meeting that the Clinical
18
Pharmacology Subcommittee had last November.
19
[Slide.]
20
Just in terms of review, this committee is
21
serving to provide expertise in three different
22
areas to this parent committee:
pharmacometrics or
23
exposure-response modeling, pediatrics, and
24
pharmacogenetics. As you see,
those were the three
25
topics that we discussed.
35
1
[Slide.]
2
Our first topic in the November meeting
3 was
a proposal by Dr. Lesko from OCPB to institute
4 End
of Phase 2a Meetings. Those are meetings
that
5 are
currently not recommended or that are currently
6 not
required by the FDA.
7
He, as well as Dr. Lee, presented the
8
FDA's perspective, and then we had three FDA
9
staffers giving us case reports where those
10
meetings may be helpful in finding optimal doses
11
early on and identifying key issues.
12
[Slide.]
13
The committee appreciated that this was a
14
pilot program that is intended to improve dose
15
findings over a few years. There
was some
16
discussion as to how we assess the success of a
17
program.
18
The committee noticed that there would be
19
additional FDA resources required to implement this
20
very program, but on the positive end, that this
21 End
of Phase 2 Meeting Program would allow
22
integration of preclinical information both in the
23 PK
and PD area and particularly to identify early
24 on
the use of biomarkers in Phase 2 and Phase 3
25
studies that may help streamline the dose finding
36
1
process.
2
The committee also felt that a meeting
3
such as this would be very useful in identifying
4 key
issues early on and discuss them between the
5
sponsor and the FDA, as well as define what we call
6
"utility" functions, which are basically measures
7 of
the potential consequences of either safety or
8
efficacy issues which are essential to come up with
9 an
optimal dose.
10
There was, as I said before, some
11
discussion as to how you would measure the success
12 of
such a program, and the committee felt that
13
probably the overriding metrics to measure the
14
success would be customer satisfaction, the
15
customer being both the sponsor, as well as the
16
FDA.
17
Possible, but more difficult to measure
18
outcome would be the need to have post-approval
19
dose changes. Again, if we can
minimize that, that
20
would indicate that there is success in this
21
program.
22
So, while the committee was in support of
23
this program, and as far as I know, it is being
24
implemented as speak.
25
[Slide.]
37
1
The second issue relating to
2
exposure-response was the issue about clinical
3
trial simulations specifically with the intent to
4
assess the liability of drug products to induce QT
5
changes which are thought to be associated with
6
fatal cardiac arrhythmias, we had Dr. Lee give the
7
introduction, Dr. Bonate from the outside review
8
modeling that he had done, clinical trial
9
simulations, and then Dr. Kenna from the FDA review
10
ongoing project within the FDA.
11
[Slide.]
12
There was a lively discussion on this very
13
topic. The committee I think still felt that the
14 QTc
correction methods, those are ways to correct
15 the
QT interval for change in heart rate, that
16
those methods are still questionable, we still
17
don't have a gold standard on that.
18
We felt that despite the trial simulations
19
presented to us, it still appears very difficult to
20
separate drug-induced changes from baseline changes
21 in
those EKG intervals.
22
There was some discussion as to what
23
constitutes a meaningful QTc change.
Right now the
24
perception is that a 6-millisecond average QTc
25
change would be relevant. There
is some concern in
38
1 the
committee or there was some concern stated in
2 the
committee that that might be too conservative,
3
however, there was acknowledgment that using
4
clinical trial simulation to get to the issue as to
5
what the QTc liability is of a new product may
6
provide a more rational risk/benefit assessment.
7
One issue that was brought up that is
8
currently not being explored is the fact that some
9
drugs, not only interact at the kinetic level, but
10
also the dynamic level, which may lead to QTc
11
changes on the PD level.
12
[Slide.]
13
The second major topic related to the
14
pediatrics component of the committee, here, we
15
reviewed the pediatric decision trees.
We had
16
several speakers. We had Dr.
Hinderling and Dr.
17
Chen giving case reports. Those
were drugs or drug
18
products that were reviewed for the pediatric use,
19
used what is a called a "pediatric decision tree,"
20
that allows PK or PK/PD studies to support efficacy
21 and
safety.
22
We had Dr. Machado giving a statistical
23
overview on what methods might be useful to compare
24
pediatric exposure-response to see whether there
25 are
any age-related differences.
39
1
Then, our committee member Dr. Kearns gave
2 his
perspective on how those studies actually are
3
being done in practice and what some of the
4
shortcomings are of the current pediatric decision
5
tree, and this was followed by Dr. Rodriguez giving
6 the
FDA experience with the decision tree that has
7
been in place for a few years.
8
[Slide.]
9
There was some discussion about the age
10
appropriateness of some of the endpoints that are
11
currently required to measure the pharmacology of
12
drugs in children, whether the endpoints are
13
related to the mechanism of action of the drug
14
and/or the pathophysiology of the disease, are
15
those meaningful endpoints and what do they tell
16 us.
17
There was some discussion, because that is
18
part of the decision tree, as to what evidence
19
supports that the disease progression in children
20 is
similar to the one in adults, which would then
21
allow it to transfer information from adults to
22
children.
23
There seemed to be consensus that
24
nonclinical information, such as data from primate
25
studies or in-vitro studies may be very useful in
40
1
supporting the pediatric decision tree.
2
However, there was extensive discussion on
3
whether there has to be extensive interaction and
4
discussion between both the clinical pharmacology,
5 the
OCPB, as well as the reviewing divisions on the
6
pediatric decision tree and its use in a particular
7
drug product area.
8
There was some discussion also on the
9
limitations of the exposure-response in terms of
10
some of the PD differences that are very difficult
11 to
be captured in the current paradigm.
12
I think there was overall an appreciation
13
that the pediatric decision tree is still
14
work-in-progress and additional updates may be
15
necessary to review or start discussing any changes
16 to
it.
17
[Slide.]
18
The last area that we discussed related to
19 the
pharmacogenomics and the metabolic drug
20
interaction area, so we had two outside speakers,
21 Dr.
Flockhart and Dr. Neuvonen talk about two
22
relatively novel cytochrome p450 isoenzymes that
23
start to emerge as part of drug metabolizing
24
enzymes, and the issue was here what is the current
25
state-of-the-art, what can FDA use as basis of
41
1
review for new incoming NDAs.
2
[Slide.]
3
There was acceptance by the committee for
4 cytochrome P4502B6, that we do have both
in-vitro,
5 as
well as in-vivo, substrates, model substrates
6
that can be used for drug interactions.
7
We don't have, on the other hand, any
8
specific clinical inhibitors, and somewhat
9
questionable in-vitro inhibitors.
On the other
10
hand, for cytochrome P4502C8, we do have both
11
in-vitro, as well as in-vivo, inhibitors, as well
12 as
substrates, so we can characterize any
13
interaction potential for cytochrome P4502C8.
14
Discussion by the committee followed that
15
went beyond the specific isoenzymes where the
16
committee emphasized that it is becoming more and
17
more essential to look at population-based clinical
18 studies
to primarily assess, not the incidence of
19
drug interactions, but their clinical significance.
20
In other words, we have enough science to
21
support the likelihood of drug-drug interactions,
22 but
we are not always sure about what the clinical
23
consequence would be or consequence would be.
24
Along the same line, the committee made
25 the
recommendation to encourage sponsors to review
42
1
databases that exist, medication-use databases, to
2
look for this very issue, what are the clinical
3
consequences of drug-drug interactions especially
4 if
you go beyond two interactions.
5
[Slide.]
6
The last topic that we
discussed related
7 to
pharmacogenomics. Again, this is an
ongoing
8
discussion that we had. In this
case, we were
9
discussing how to integrate that in the drug
10
development and what kind of labeling may be
11
necessary to reflect information collected during
12 the
development process.
13
We had committee member Dr. Flockhart and
14 Dr.
Relling give their academic, as well as
15
clinical, perspective, and Dr. Hockett give the
16
industry perspective.
17
[Slide.]
18
To summarize the committee discussion, I
19
think there was acceptance of the fact that we need
20
additional population-based studies meaning
21
large-scale studies to look at the prevalence for
22
some of the rare genetic polymorphisms, in other
23
words, for some of those polymorphisms that may be
24
important, we do not know how many patients have
25
those specific genotypes.
43
1
There was recognition that we do have or
2 at
least start to emerge having a lot of
3
mechanistic and quantitative understanding that is
4
necessary for labeling.
5
In other words, we collect a lot of
6
information and we know a lot about how likely some
7 of
those pharmacogenetic differences are and what
8 the
kinetic or dynamic consequences are.
9
The discussion then really focused on what
10 is
the impact as far as risk/benefit is concerned,
11 in
other words, how do we translate changes in drug
12
levels or change in the pharmacology of the drug,
13 how
do we translate that into safety and efficacy
14
information.
15
There was, shall we say, a lively
16
discussion of how to label pharmacogenetic
17
information in drug package insert, and I don't
18
think there was any consensus.
19
We had experts telling us we need to label
20
very extensively, on the other hand, clinicians
21
were concerned about overloading information that
22 is
not being used by the ultimate consumer, and
23
there was recognition that pharmacogenetics or
24
pharmacogenomics is going to be different from some
25 of
the other clinical covariates in the sense that
44
1 it
has multidimensional nature, in other words,
2
there are lots of different pharmacogenetic
3 polymorphisms that may be relevant for a
given drug
4
product.
5
I would be happy to entertain any
6
questions that you may have.
7
DR. KIBBE: Okay. Jurgen will be with us,
8 so
if you want to ask questions later, if topics
9
come up that we need to get back to him on, we can.
10
Thank you.
11
Now, I know you are fumbling through your
12
things looking for the slides for the next speaker,
13 but
there aren't any, which gives us great hope
14
that it will be a short and direct presentation.
15
Dr. Hussain.
16
Parametric Tolerance Interval Test for
17 Dose Content Uniformity
18
DR. HUSSAIN: No, I do not have
slides for
19
this part of my introduction. The
topic that will
20 be
discussed as a proposal to you is that of
21
parametric tolerance interval test.
22
As we have discussed this several times
23
with you, in particular at the last meeting, in the
24
previous meeting that we had, the challenge is how
25 do
you move forward with adopting a more rigorous
45
1
scientific, statistically sound approach to dose
2
content uniformity of inhaled products.
3
We believe that parametric tolerance
4
interval test that is being proposed by IPAC-RS is
5 an
improvement over the current method, and we
6
would like to sort of move forward in sort of
7
resolving some of those issues which have lingered
8 on,
and sort of adopting it as soon as possible.
9
But the challenges are not trivial, and I
10
tried to sort of summarize those challenges to you
11 in
the memorandum along with the paper that we
12
wrote.
13
We felt that in order to move this process
14
faster and move it forward more quickly, the
15
proposal to you is that we will form a working
16
group under this advisory committee.
17
This working group will report to you with
18
their findings and provide a way forward to
19
resolving the issues that have lingered on for
20
three years, and come up with a very well
21
structured process to resolve in a timely fashion.
22
So, the proposal is a very straightforward
23
proposal that this working group will report to
24
you, and you will define the goals and objectives
25 for
this group, and you will define also the
46
1
timeline for this group, and the proposal will be
2
presented by Bob O'Neill, who is going to head for
3 FDA
working group members.
4
Bob.
5 Moving Forward -- An Approach for
Resolution
6
DR. O'NEILL: Good morning.
7
[Slide.]
8
My name is Bob O'Neill, and as I indicated
9
earlier, I am the Director of the Office of
10
Biostatistics, and Ajaz and Helen have asked me to
11
chair this group, which Ajaz has indicated is going
12 to
be reporting to you all.
13
This is the process for coming to
14
resolution on what you know to be a discussion that
15 has
been going on at least for three years under
16 the
specifications for delivered dose uniformity
17 for
inhaled and nasal drug products.
18
[Slide.]
19
I am going to be proposing how we are
20
going to be going about doing this and asking for
21
your advice and concurrence, so we can move forward
22 on
this.
23
So, what we have thought about, and we
24
have met several times with the IPAC-RS group, and
25
this is the proposal. We will
have a joint working
47
1
group under this particular committee, and it will
2 be
populated by senior representatives from FDA and
3
from the Oral and Inhaled Nasal Drug Product
4
industry, and that is mainly the IPAC group that we
5
have been working with.
6
[Slide.]
7
The folks from FDA, I will get into the
8
names in a moment, but essentially are representing
9
sort of the clinical risk side of the house, the
10
statistical side of the house, the generic drug
11
side of the house, and the Office of New Drug
12
Chemistry side of the house, so all the major
13
players in terms of how this particular solution
14 impacts
the way we go about doing business.
15
This particular proposal is essentially a
16 way
forward, so that we have a defined process with
17
identified objectives, with identified ways of how
18 we
are going to communicate with each other, in
19
terms of the mechanism, some timelines, some
20
milestones, and how are we going to get some
21
resolution on some of the issues that might be sort
22 of
sticky or still needing further discussion.
23
So, the overall working group objective is
24 to
agree on a mutually acceptable parametric
25
tolerance interval test for delivered dose
48
1
uniformity, and these are the folks, and if they
2 are
in the room, I would ask them to stand up.
3
On the lefthand side are the FDA folks.
4 It
is myself, Dr. Chowdhury, I believe Badrul is
5
here. He is the Pulmonary
Division Director.
6
Moheb Nasr, I believe is out of the country, you
7
probably know him. And Lawrence
Yu, I don't know
8 if
Lawrence is here--there he is, and he is the
9
Director for Science in Office of Generic Drugs.
10
On the industry side, I think Michael is
11
here, Michael Golden from GlaxoSmithKline. Kristi
12
Griffiths, I don't know if she is here, from Eli
13
Lilly. Bo Olsson from AstraZeneca.
Dar Rosario
14
from Aradigm. Dennis Sandell from
AstraZeneca
15
also. We have met with these
folks and we plan on
16
meeting in the future, and I will go through the
17
timeline.
18
[Slide.]
19
So, just to reiterate, the objective of
20
this working group is to develop a mutually
21
acceptable, standard DDU specification, both the
22
test and the acceptance criteria, for these
23
products with a proposal to come back to you folks
24 by
the end of this year, by the end of 2004.
25
[Slide.]
49
1
So, the process that we are going to
2
follow is pretty much trying to get the
3
communication and the coordination of this effort,
4
which is not going to be trivial, straight among
5 all
of us.
6
We have identified that we will have a
7
project manager that will help us as a working
8
group stick to agendas, minutes, meeting materials.
9 We
plan on having monthly meetings at FDA beginning
10 in
May.
11
The first one is probably in a few weeks,
12 and
in May, what we will plan to do is to review
13 the
feedback that you all give us today in terms of
14
your blessing and what other suggestions you might
15
have for how we would fine-tune this particular
16
process.
17
We are going to need to rely on working
18
groups within the industry and within the FDA to
19
further deal with the statistical issues here, the
20
clinical issues, the CMC issues, and whatever else
21 is
on the plate, so there is likely to be some
22
technical projects that will be assigned to folks,
23 and
the leadership and the project management of
24
those particular projects will be overseen by the
25
folks on the working group.
50
1
[Slide.]
2
So, again, just to reiterate the timelines
3
and the milestones, we expect to
have a status
4
report back to you folks in the
fall, in the
5
meeting in the fall, in October, and hopefully to
6
submit recommendations to you by the end of 2004
7
that you can act on and come back to us on.
8
[Slide.]
9
Here is where we think we are to date.
We
10
have discussed these issues at length and here is
11
what we think we have reached consensus on.
12
That the parametric tolerance interval
13
approach is an improvement on the current test. It
14 is
a concept that requires refinement and further
15
development to address the regulatory requirements.
16
There are still things that need to be fine tuned.
17
We believe that there has been a lot of
18
work, productive work, a lot of understanding, but
19 it
is time to move forward and come to closure
20
particularly on this particular test.
21
So the working group is formed to devote
22 the
necessary time and the resources to get this
23
thing done, and that is through review of
24
additional data analyses, especially some of the
25
appropriate statistical procedures.
51
1
[Slide.]
2
We also recognize that there is some stuff
3
hanging out there that needs consensus.
You have
4
probably seen a presentation and heard about a
5
presentation with regard to the different operating
6
characteristic curves, the parametric tolerance
7
interval test versus sort of the zero tolerance
8
test, and there is a gap that essentially is the
9
difference between the producer and the consumer
10
risk, and it sort of differs in the middle over
11
what you might assume to be the standard deviation
12 of
some of the measurements.
13
That is essentially where a lot of the
14
discussion has been. Much of the
discussion has
15
been around what the performance characteristics
16 are
of the different tests under assumed scenarios.
17
Another way of saying assumed scenarios is the
18
simulated data, so if this, then that.
19
So, if the data were to perform this way
20 or
lay itself out this way, then, this is what the
21 operating
characteristics of that particular test
22
procedure are.
23
So, we are actually also interested in
24
seeing what real data is, so there is a number of
25
issues with regard to actual data that is not in
52
1 our
hands, not in FDA's hands, which would lead us
2 to
say, well, how many situations are there where
3 the
standard deviations start to push out to 12,
4 13,
14, 15, because those are the areas where you
5 may
be wanting to have a little more information
6
because if you are not in the symmetric situation,
7
your outliers are going to be where your problem
8
cases are.
9
So, there is some more work to be done in
10
this area, so talking about that and marrying both
11 the
zero tolerance interval concept with the
12
parametric tolerance interval idea is essentially
13
where the statistical details of the test are
14
likely to be focused over the next few months.
15
Obviously, this issue of the applicability
16 to
non-normal distributions, asymmetric bimodal
17
distributions, which essentially may be very much
18
characteristic of manufacturing processes of, you
19
know, large and small particles, and things like
20
this, which is not an unusual statistical scenario
21
when you have mixtures of populations, so that is
22
from the statistical perspective.
23
[Slide.]
24
The next steps are to ask
you folks to
25
endorse this idea or to suggest some refinements to
53
1
it. We will come back to you with
a status report
2 as
to where we are in October, and the working
3
group is planning to submit recommendations to you
4 all
by the end of this calendar year.
5
With that, I think I am done. I
would be
6
willing to take any questions, and I think anybody
7 on the working group would also be willing to
chime
8 in.
9
Committee Discussion and Recommendations
10
DR. KIBBE: We have time now for
11
questions, it's on our schedule, so ask questions.
12
DR. SINGPURWALLA: Well, just out
of
13
curiosity, what is the DDU test?
14
DR. O'NEILL: Delivered dose
uniformity
15
test. It is essentially a
measurement of, it's
16
content uniformity, how much of the dose is
17
delivered in, let's say, a spray or these nasally
18
inhaled products, so it's a matter of if this was a
19
pill, you would be crunching it up, you would be
20
looking at what its content is, you would have a
21
measure of that, and the test is essentially that
22 you
agree what the goalposts are for an acceptable
23
amount of variability for the active ingredient,
24 and
if it's in that zone, it's acceptable; if it's
25 not
in that zone, it is not acceptable, so it's a
54
1
variant.
2
That is the whole concept behind the
3
delivered dose uniformity, that the product has to
4
have some consistent uniform characteristics to it.
5
DR. SINGPURWALLA: So, how would
it differ
6
from the parametric tolerance interval?
7
DR. O'NEILL: Well, first of all,
it
8
differs in a number of ways. I
don't want to go
9
through the test, that there has been a
10
presentation on this, and there is a lot of
11
background stuff on this.
12
The key difference between the zero
13
tolerance is it's a zero/1 kind of thing, it's
14
either in or out, and it doesn't take the standard
15
deviation into account.
16
The parametric tolerance interval approach
17 is
probably, assuming that you have something close
18 to
normality, and it is essentially basing the test
19
both on the estimate of the mean and the estimate
20 of
the standard deviation, and then depending upon
21 the
combination of both of those guys, it is
22
essentially a zone of equivalence, but the
23
distinction between the two tests is one sort of a
24
zero/1, you are either all in or all out, but it
25
doesn't estimate the standard deviation.
55
1
The work that has been done on the
2
parametric tolerance interval approach
3
statistically is intended to be a more powerful,
4
more precise, take more of the information into
5
account.
6
DR. SINGPURWALLA: So, would you
say that
7 the
DDU test is not a statistical test, it has no
8
statistical basis?
9
DR. O'NEILL: No, I would not say
that at
10
all. In fact, both of them have
statistical bases.
11 In
fact, the zero tolerance test is essentially the
12 USP
test that is used for all content uniformity,
13 it
is a variation on that.
14 Take 10, see whether they are in the
15
limits or out of the limits, if not, take another
16
20. If they are in the limits or
out of the
17
limits, and you are done, up or down.
That is what
18 the
test has been for years.
19 What this is, is essentially to
say,
20
well, I am not using all the information, I am not
21
finding out actually what the variability of the
22
process is, so I want to get some handle on what
23 the
standard deviation of the process is, so I want
24 to
estimate that also, and I also want to estimate
25
what the mean is.
56
1
So, if you were to back up and sort of
2
look at this within the mainstream of process
3
control, you sort of want to look at where you are
4 in
the standard deviation world, where you are in
5 the
mean target close to what the center of the
6
distribution is.
7
So, both of these are statistical in the
8
sense that they have probabilities of consumer risk
9 and
regulatory risk, but it is that part of it that
10 is
the statistical aspect of it.
11
DR. SINGPURWALLA: So, if I were
to
12
understand what you are saying, the DDU test seems
13
like a binary test, it's a sequential binary
14
process.
15
DR. O'NEILL: Well, what we are
talking
16
about, we are talking about the parametric
17
tolerance interval test versus what is--I don't
18
know what its best name is--but it would be like
19 the
zero tolerance interval test. That test
is
20
binary. The other one is--
21
DR. SINGPURWALLA: Is not binary.
22
DR. O'NEILL: --is not binary. It takes
23
more of the information into account.
That is the
24
conceptual idea.
25
DR. KAROL: Could you tell me how
much
57
1
real data you have and what is the source of the
2
real data?
3
DR. O'NEILL: Well, we have, our
folks, I
4
know that there are folks maybe in the audience who
5
have looked at data that we have from the industry,
6 but
it is not necessarily the data that is all the
7
data.
8
I mean what we have is data that is
9
submitted to us in applications and in annual
10
reports, and often that is data that has already
11
been screened in the sense that it either passes or
12 doesn't
pass, so in some sense, we are seeing data
13
that is less variable than the data that these
14
tests are intended to apply to uniformly.
15
I believe that is where our comfort level
16 is
in terms of trying to understand how much
17
variability is in the data, and I think it's a
18
conceptual thing getting back to the way Helen
19
talked about.
20
For years, for years, I think the process
21 was
let's set the goalposts and then see whether we
22 can
manufacture it to fit the goalposts as opposed
23 to
the other way around, sort of saying what is the
24
process capability and then fix the goalposts for
25 the
process capability.
58
1
Under continued process improvement, the
2
idea is to be closer to the target mean and to be
3
closer and tighten down your variability. It may
4 be
if you can't do any better, that's what you are
5 left
with.
6
So, our situation is understanding that,
7 and
what we are seeing now is I believe, if I am
8 not
speaking for our chemists, our folks are seeing
9
relatively tight standard deviations in the 5, 6, 7
10
area, and the idea that there could be some
11
standard deviations that are hanging out in the 12,
12 13,
14 area is how come. We are not
necessarily
13
seeing all of that.
14
So, we want to see a little more data
15
along those lines. So, that is
sort of
16
conceptually where the gap is in terms of trying to
17
move transitionally from the current test into a
18
test that we believe has a lot more merit for
19
several reasons.
20
One, it captures better a handle on the
21
variability of the data, and, secondly, you should
22 be
rewarded for taking more samples than less
23
samples. So, this test needs to
reward you for
24
having better estimates of what your variability is
25
rather than less. That is another
conceptual part
59
1 of
this.
2
DR. MEYER: I might be mistaken
because I
3
don't normally read the USP, but it seems from my
4 recollection
there are some tablet products that
5
have a specification for variability, as well,
6
warfarin being an example, where you do 10, then
7 you
do 20, but you also look at standard deviation
8 or
coefficient of variation as some marker for
9
approval or not.
10
Is that correct?
11
DR. O'NEILL: Ajaz.
12
DR. HUSSAIN: Right, I think,
Marv, you
13 are
right, in the sense the traditional approach,
14 in
the pharmacopeial approach, which are market
15
standards, and they were never intended to be
16
release standards, and that is the purpose they
17
serve, are to maintain the market standard.
18
In the case of tablets and solid dosage
19
forms, you have a non-parametric approach to that,
20
and, say, you have your goalposts 85 to 115 for 10
21
tablets, and if one is outside that, you go to 75
22 to
125 with 20 additional ones.
23
For those, you have an estimate of
24 standard
deviation. I think it's 6.6 person at
the
25
second stage, so you have to meet
that.
60
1
The test we have for dose content
2
uniformity or delivered dose uniformity for
3
inhalation products right now, the FDA guidance
4
doesn't have a value of standard deviations. It
5
simply says take, if it's 85 to 115, if one is
6
outside that, take 20 more, and they all have to be
7
within 75 to 125.
8
So, the term "zero tolerance" actually is
9 not
really a meaningful term, and I think we
10
discussed that at the previous committee, but if
11 you
really look at it, Jurgen had one set of
12
comments at the end of that meeting, and our
13
statisticians there had a very different set of
14
comments on that, so we were very divided on that,
15
because zero tolerance is for that sample, and that
16 is,
in my opinion, a big hindrance to continuous
17 improvement
because it forces industry to do only
18 30
tests.
19
If they do more, they are at risk, so that
20 is
not conducive to PAT, that is not conducive to
21 the
21st century process that we want to move
22
forward, so this actually is a model or the
23
framework for what we would like to do for all
24
specification, because clearly, the compendia,
25
there is no movement.
61
1
I don't see much movement in the compendia
2 to
change that, so we will have to move forward and
3
change that, because if the compendia don't change
4
that, they are going to be hindrance to PAT and
5
everything else that follows.
6
DR. BOEHLERT: Just as a follow-up
to
7
that, I believe under ICH, the compendia are
8
looking at harmonizing general chapters, and one of
9 the
ones they are looking at is content uniformity
10 and
should there be a tie-in somewhere with that
11
group and what they are looking at and what they
12 are
doing, so you don't go two separate ways in two
13
separate directions.
14
DR. HUSSAIN: I agree, but
compendia are
15
still a market standard, they are not a release
16
standard, so from a regulatory perspective, that
17 has
always been the case.
18
DR. BOEHLERT: That has always
been the
19
case.
20
DR. KIBBE: Tom.
21
DR. LAYLOFF: I was going to say
also
22
there is a market standard in the way--you end up
23 in
a contradiction if you test the whole lot, it
24
will always fail, because of the standard
25
deviation, so you can't really do that.
62
1
But in the regulatory laboratory, what we
2
used to do is if we found one out of limits, then,
3 we
would submit it for check analysis, and if it
4
passed check analysis, then, it was okay. So, you
5
sort of got around that contradiction in the limit
6
setting.
7
DR. KIBBE: Anybody else?
8
Is there anyone on the committee who
9
thinks that moving forward is not necessarily the
10 way
to go? Is there something that we need
to
11
discuss, because they are essentially asking us to
12
say, well, yeah, we need to move forward and let's
13 get
the results by the end of the year?
14
DR. SINGPURWALLA: Do we have to
do this
15
right now?
16
DR. KIBBE: We are not going to
decide on
17
which tests to do right now. We
are just
18
supporting the concept of having the working group
19
move forward and give us a report.
20
DR. SINGPURWALLA: But one of the
things
21
they wanted is recommendations .
22
DR. O'NEILL: No, I don't think
so. We
23 are
just asking you to endorse the idea of moving
24
forward and having this group, and we will come
25
back to you with a report.
If you don't like it,
63
1 you
can say go do more.
2
DR. SINGPURWALLA: I am sorry,
you said
3
suggest refinements in your talk, I made a note of
4 it,
so do you want the refinements now or later on?
5
DR. O'NEILL: No, we don't.
6
DR. SINGPURWALLA: So, you don't
want
7
refinements.
8
DR. O'NEILL: No, it's very high
level,
9 not
detail oriented feedback that we would like
10
from you right now.
11
DR. SINGPURWALLA: Because I would
like to
12
suggest refinements, but not at this minute.
13
DR. O'NEILL: I am sure we would
be very
14
interested in your refinements, and, in fact, I
15
would certainly be interested in speaking with you
16
outside of the meeting in terms of getting some
17
additional ideas on this particular test, because
18
again, this is a working group that is under the
19
umbrella of this committee and essentially is
20
coming back to the committee on behalf of the
21
committee saying what do you think, because the
22
committee is the one who is going to give the
23
recommendations to the Agency.
24
So, if you don't like the recommendations,
25
then, it is totally within the committee's
64
1
responsibilities and rights to say, you know, that
2 is
not what we had in mind, or that's not what we
3
think is right.
4
DR. KIBBE: Let me get at some of
this a
5
little bit. We have, I think, a
tentative schedule
6 to
meet in October, and for you, the working group,
7 to
have your best shot prepared for us to look at
8 and
give you feedback on, right?
9
DR. O'NEILL: Yes, and it's not
that we
10
haven't thought this isn't ambitious either, but
11
that's what we are trying to
work on.
12
DR. KIBBE: Is it reasonable for a
member
13 of
this committee to forward suggestions to you in
14 the
interim and then have you incorporate them in
15 the
working group? If you have some things
that
16 you
would like to think through and then--
17
DR. SINGPURWALLA: Honestly, I was
18
intrigued by the comment made that we invite
19
suggested refinements, and for me to suggest
20
refinements, I need to have a better appreciation
21 for
exactly what is going on.
22
DR. O'NEILL: I hear what you are
saying.
23 I
guess maybe that was meant in terms of
24
refinements to the process. Part of this is the
25
process, and part of this is the content that the
65
1
working group will be dealing with, and the working
2
group already has essentially a proposal that they
3
have been reacting to from IPAC-RS that has been in
4 the
works for a number of years, and it is that
5
that is trying to be refined, those ideas are
6
trying to be refined in the context of how do we
7
understand what is currently sort of the operating
8
characteristic curve of the current way we do
9
things versus a new proposed way of doing things,
10 and
are they achieving where we want to be as a
11
committee.
12
I think that is the sense of the
13
refinements.
14
DR. SINGPURWALLA: So, if the
endorsement
15
that you seek is for the process, and not for the
16
inner workings of the process, I have no comments,
17 go
ahead, but if it is for the workings, then, I
18
would like to think about it.
19
DR. KIBBE: I believe we are
looking for
20
moving ahead on the process right now.
21
DR. O'NEILL: That is what we are
seeking
22
from you, yes.
23
DR. KIBBE: What I hear my
colleague
24
saying is that he would like to have some input on
25 the
actual workings of the committee, with the
66
1
thought process of the committee, and that if we
2
could find some way to do that, to accommodate that
3
situation within the budget constraints of the FDA,
4 it
would be useful.
5
It always is good for a subcommittee or a
6
working group of ours to have somebody from here to
7
carry water for us. You might get
yourself into
8
more work than you thought you were going to get
9
into.
10
Anybody else? Jurgen.
11
DR. VENITZ: I am obviously in
favor of
12
moving forward, but I would like to give maybe
13
somewhat of an unwanted recommendation, not
14
necessarily a refinement.
15
That is, when I look at the objectives of
16 the
working group, they are basically, primarily
17
looking at the statistical properties of the test.
18
I am recommending the group for having
19
information on it, and I would encourage the
20
committee to also, the subgroup, I guess, the
21
working group, to also look at the clinical
22
significance, in other words, in my mind, we talked
23
about that last time, the clinical use is part of
24 what
risk-based manufacturing is all about.
25
So, for example, it may be very different
67
1
whether you are comparing inhaled insulin release
2 to
inhaled topical steroids, and I would like for
3
that to be discussed as part of the working group.
4
DR. O'NEILL: I hear you. Maybe I went
5
through this a little too fast.
If you look at the
6
constitution of the working group, Dr. Chowdhury is
7 our
clinical input on that, so that has been
8
recognized, and that is why he is on the working
9
group, to essentially put, as an overlay, the
10
clinical risk structure on this, recognizing very
11
much it might be product-specific, so that is his
12
role.
13
Lawrence Yu's role is also looking at this
14
from, let's say, the generic drug implication, so I
15
think the working group has been put together
16
primarily to be relatively broad-minded.
17
The statistical component of this is only
18 one
of multiple dimensions to this, but it is
19
critical to understanding where we are in terms of
20 the
only thing that is not moving right now, which
21 is
the test that is on the table.
22
DR. KIBBE: Pat, go ahead.
23
DR. DeLUCA: Since this committee
is going
24 to
be reporting back to this group, I am just
25
wondering why a member of this group wasn't put on
68
1
that committee, and it sounds like Nozer could have
2
some real input into it, as well as being a link to
3
this committee. There may be some
reason why you
4
didn't do that, but I would certainly consider
5
that.
6
MS. WINKLE: It certainly is an
option.
7 The
way that this group is set up is basically a
8
fact-finding group for the advisory committee, to
9
give them the facts and the information that they
10
will need to help make a recommendation on this
11
test and how we want to move forward with it, but I
12
think that it would be very helpful to have some
13
input from Nozer.
14
I think that he has some knowledge and
15
some understanding and there is nothing that
16
prohibits us from doing that, but we tried to set
17 it
up as an independent fact-finding group for the
18
advisory committee.
19
DR. SINGPURWALLA: By the way, I
just want
20 to clarify that I didn't raise the question to
21
thrust myself into this arena. I
was honestly
22
asking a question, and since the matter has been
23
raised by my colleague on the clinician, I would
24
like to suggest that a Bayesian be on this
25
particular group.
69
1
DR. O'NEILL: We will certainly be
2
listening to you. If you want to
get into that
3
discussion, we could, but one of the critical
4
discussions we have been having right now is
5
assumptions versus data, and Bayesians are heavy on
6 the
assumptions, but you have to have the data to
7
support the assumptions, the game we are in, in the
8
regulatory game we are in, and that is why we are
9
trying to sort of get some sense of what does the
10
waterfront actually look like, because it is very
11
important to the behavior of the characteristics of
12
this test.
13
DR. KIBBE: It is always fun to
have
14
statisticians discussing statistics.
15
Do we have any other questions?
16
Seeing no one's hand or little button lit
17 up,
I want to thank you very much. We are
looking
18
forward to a very informative and useful report in
19
October.
20
My schedule says that we are supposed to
21 be
talking until 10:15, and we could either take a
22
break now or if Ajaz promises to get finished in
23
time for a break, we could move forward.
What is
24
everyone's pleasure? Naturally,
the Bayesian wants
25 to
break.
70
1
[Laughter.]
2
DR. KIBBE: I will give you all 15
minutes
3 and
then we will have Dr. Hussain.
4
[Break.]
5
DR. KIBBE: Why don't you go ahead
and
6
start, Ajaz.
7
Process Analytical Technology (PAT) - Next Steps
8
DR. HUSSAIN: Thank you.
9
[Slide.]
10
What I would like to do today is to give
11 you
a brief progress report on the PAT initiative
12 and
have three speakers.
13
[Slide.]
14
I will present a brief history to recap
15 how
we got here, current status and next steps.
16
There are three topics that we want to share with
17
you, finalizing PAT guidance, training and
18
certification. Chris Watts will
make that
19
presentation.
20
What we are doing with respect to
21
standards development. Ali Afnan
will talk about
22
that.
23
A topic that we have discussed twice with
24
you, but we thought we would sort of bring some
25
closure to that, what we have done with rapid
71
1
microbial methods and how that has been a part of
2
PAT. Bryan Riley will talk to you
about that.
3
What we are hoping is, we have not really
4
posed any questions, this is more of a progress
5
report, status report, and we are moving forward,
6 but
if there is anything that you think we need to
7
consider, please share this with us.
8
The questions you might want to consider -
9 are
we on track? Are there any
recommendations for
10
improving how we have approached PAT and how we
11
might want to approach PAT in the future?
12
[Slide.]
13
The aspect that I often share is I think
14 the
PAT thought process has been in the Agency for
15 a
long time, and, in particular, a focal point for
16 the
discussion occurred in October of 1993.
I was
17 not
at FDA at that time, but Tom Layloff and others
18 in
St. Louis had organized a Symposium on
19
Pharmaceutical Process Control and Quality
20
Assurance by Non-traditional Means.
21
The information I have about that is a lot
22 of
the focus became on near IR, and a lot of the
23
focus tended to be on endproduct testing although
24 the
title was process control, and the discussion
25
that led to sort of a very negative view of near IR
72
1 and
some of this technology came from FDA saying
2
this cannot be USP methods, therefore, cannot be
3
regulatory methods, which is probably more blunt,
4 Tom
will correct me if I am wrong.
5
So, I think that was really an unfortunate
6
aspect because from an FDA perspective, a lot of
7
progress did not occur because of that.
8
Tom and I spent a lot of time together
9
thinking about this, and we saw this as an
10
opportunity. It was more of a
discussion between
11 an
analytical chemist and an industrial pharmacy
12
type, so we were putting our heads together and we
13
made a presentation in the year 2000, the
14
Millennium Conference in San Francisco.
I will
15
just share some slides on that with you.
16
Another meeting which was very important
17 in
the evolution of this process was the new
18
technology meeting of Royal Pharmaceutical Society
19
entitled Process Measurement and Control. I
20
actually met Ali Afnan and many other people who
21
were then associated with the PAT at that meeting.
22
[Slide.]
23
The aspect I think which was important is
24
this was a presentation that Tom and I did together
25 at
FIP meeting. Tom had left FDA and was
part of
73
1 the
USP at that time. The title was Advanced
2
Quality Control of Pharmaceuticals: In-line Process
3
Controls.
4
If you look at the outline, what we talked
5
about then was pharmaceutical product development
6 and
manufacture: Building Quality In, and sort of
7
design and specifications, how you approach that.
8
We looked at modern in-line controls,
9 potential advantages over traditional
controls, a
10
better approach for "building quality in," and
11
talked about the need for accelerating industry and
12
regulatory acceptance of modern in-line controls.
13
That was the thought process before we coined the
14
term "PAT," and so forth.
15
[Slide.]
16
In many sense, if you look at the cartoon
17
there, that was the art of pharmacy manufacturing
18 to
the science of pharmaceutical manufacturing is
19 how
did we do granulation endpoint. We reach
in
20 the
bowl, grab a handful of granules, and look how
21
they crumble, and then decided the granulation
22
endpoint was reached, so we wanted to move from the
23 art
to more of a science-based approach.
24
Our part of the PAT looked something like
25
this, so if you look at that other cartoon there,
74
1
that is how we saw it in 2000, this is what PAT
2
might be.
3
[Slide.]
4
I think one of the critical meetings that
5 I
attended was a far more technical conclave in
6
North Carolina. I happened to
walk into that
7
meeting and G.K. Raju from MIT was talking about
8 it,
and that was a chance meeting that really
9
provided us some of the critical information
10
because I think without that, Tom and I could not
11
have made any points in 2001.
12
What the CAMP consortium, the MIT
13
consortium helped us was to really put a value to
14
this thought process, and based on that, we made a
15
presentation to the advisory committee, Vince Lee
16 was
the chair then, is to initiate public
17
discussion on application of process analytical
18
chemistry tools in pharmaceutical manufacturing.
19
You gave us strong support to move
20
forward. You recommended that we
form a PAT
21
Subcommittee. We also, at that
same meeting,
22
related discussion on Rapid Microbial Testing,
23
however, we did not discuss this further at the
24
advisory committee, we had these discussions at the
25
subcommittee, and that is the reason I brought
75
1
Bryan Riley to come back and share with you that
2
discussion again.
3
[Slide.]
4
But at the same time, I think Helen and
5 Dr.
Woodcock, we were discussing this, we felt this
6 was
much bigger than just an OPS issue, it had to
7 be
an FDA issue, so we took this to the FDA Science
8
Board, and Dr. Woodcock presented that as emerging
9
science issues in pharmaceutical manufacturing.
10
We actually invited--I am not going to go
11
through all the slides, but just to sort of
12
illustrate the key presentations that occurred--one
13 was
the opportunity for improving the efficiency
14
from G.K. Raju and then Doug Bean from
15
PriceWaterhouseCooper, and we had industry
16
colleagues from Pfizer who really came and helped
17 us,
saying that Pfizer has adopted a "Don't Use"
18 and
"Don't Tell" approach.
19
That is the industry approach is to not to
20 use
new science and new technology because of
21
regulatory uncertainty, or if it is needed, they
22
will use it, but then they will do something for
23 the
regulators to say here, this is what you want,
24 but
we will control the process this way.
25 So, we felt that was undesirable from
a
76
1
public health perspective, and we wanted to move
2
forward to facilitate introduction of PAT, and we
3
coined the term PAT. So, we got a very strong and
4
unanimous endorsement from the FDA Science Board to
5
move forward. In fact, the
Science Board also said
6
that they would like to talk and give seminars on
7 it,
but they have not, but we did give them
8
updates.
9
[Slide.]
10
Taking the recommendations of the advisory
11
committee, this committee's recommendation. we
12
issued a Federal Register Notice to invite people
13 to
participate on a PAT Subcommittee.
14
So, we got people to apply. We
selected
15
those individuals and we formed a PAT Subcommittee.
16 We
brought it back to this advisory committee to
17 see
whether the charter for the subcommittee is
18
acceptable.
19
You gave us valuable recommendations.
We
20
formed the subcommittee, and we had three meetings
21 -
October, June, and February. Tom Layloff
served
22 as
the acting chair for the subcommittee.
23
[Slide.]
24 The subcommittee moved so rapidly we did
25 not
have an opportunity to remove the word "Acting"
77
1
from these names, so while they were acting, the
2
work was done, so we never finalized their
3
positions.
4
Dr. Kibbe, now the current chair of this
5
committee, took the responsibility for PAT
6
Applications Benefits Working Group.
Judy
7
Boehlert, who is the chair for Manufacturing
8
Committee, took the lead for Product and Process
9
Development Working Group.
10
Leon Lachman focused on Validation.
11
Dr. Koch, who is now on the advisory
12
committee, chaired the Working Group on PAT
13 Chemometrics.
14
So, these working groups provided us
15
information, feedback to sort of help create a
16
framework to write this guidance.
17
[Slide.]
18
We also, in parallel, were discussing this
19 further
at the FDA Science Board, and the key
20
aspect was the PAT initiative was just a starting
21
point to what was to follow, the 21st Century
22
Initiative, and so forth.
23
So, we took this discussion further to the
24 Science
Board, and the second Science Board
25
discussion was very important.
There was a topic
78
1
that Dr. Woodcock herself discussed, and that was
2
actually something similar to what we had the
3
discussion on parametric tolerance interval test,
4
because the current regulatory system and the
5
current pharmacopeial system is such that actually
6
does not promote continuous improvement, it
7
actually penalizes people for doing more testing,
8 and
therefore it had to change.
9
So, we had to bring the concept of
10
research and moving away from the current mentality
11 of
75 to 125 type thinking, the market standard
12 type
thinking, so we had to build that consensus,
13 and
we got strong endorsement from the FDA Science
14
Board to move forward also on that aspect.
15
The other presentation, which is very
16
important to remember, is that of Dr. Ray Sherzer
17
from GlaxoSmithKline speaking on behalf of CAMP,
18 and
the thing that he pointed out, that there are
19
many barriers, we need a paradigm shift, and that
20
paradigm shift is necessary because the barriers
21 are
cultural, organizational, historical.
22
The challenges are not technical, the
23
technical knowhow exists. The
scientists can do
24
this, but the barriers are significant cultural
25
barriers and organizational barriers, and we could
79
1
relate to that, because we had the same barriers
2
in-house at FDA.
3
[Slide.]
4
As we were building the PAT team process,
5 and
you will see a lot of the thought processes
6
that Helen expressed in terms of the desired goal
7
that OPS wants to move in, this becomes a model or
8 the
pilot project for a lot of the things we have
9
done.
10
So, we had to build a PAT team for
11
reviewers and inspectors and compliance officers,
12
because this was the engine for success.
We had to
13
think very carefully about this because we have a
14
long history of turf issues. We
don't talk to the
15 field,
the field doesn't talk to us type of
16
mentality, or this is my issue, field keep away
17
type of thing.
18
[Slide.]
19
So, we actually started a team building
20
exercise, so starting with a definition of team, a
21
team is a group of interdependent individuals with
22
complementary skills who are organized and
23
committed to achieving a common purpose, applying a
24
common process, and sharing a common destiny.
25
Now, I think we clearly have worked on No.
80
1 1
and 2, we haven't really worked on No. 3 yet, but
2 the
importance of this is the quality of the
3
results we expect from the regulatory assessment,
4
review, or inspection really depend on the quality
5 of
relationship between the reviewer and
6
inspectors, and the quality of the relationship
7
defines quality of thinking, and the quality of
8
thinking defines quality of action that leads back
9 to
the quality of results we expect.
10
So, this is really a complex issue and
11
that has to be dealt with very carefully.
12
[Slide.]
13
We started the PAT process with three
14
organizations: our colleagues in
Office of
15
Regulatory Affairs, which are the GMP inspectors,
16
Center for Drugs, and Center for Veterinary
17
Medicine.
18
The Center for Biologics chose not to be
19
part of this, and we will discuss that further this
20
afternoon whether they wish to join us or not.
21
So, we formed a PAT Steering Committee,
22
again reflecting all the different organizations.
23 We
formed a PAT Review and Inspection Team, and we
24
actually recruited a small group, Raj Uppoor, Chris
25
Watts, Huiquan Wu, and Ali Afnan to come and join
81
1
OPS, so we had a very successful recruitment
2 process. We actually got Ali to take half the
3
salary to come to work for FDA, and he did.
4
We actually put a PAT Training and
5
Coordination Team, and the training was critical.
6 One
of the critical aspects of the PAT Subcommittee
7 was
developing a curriculum for training, and then
8 we
partnered with three schools: a School of
9
Pharmacy, a School of Engineering, and a School of
10
Chemistry to bring this process together, all three
11
National Science Foundation Centers for Excellence,
12
Center for Process Analytical Chemistry,
13
Measurement Control Engineering Center at
14
Tennessee, and Center for Pharmaceutical Processes
15 at
Purdue.
16
So, we brought the groups together and the
17
training occurred, but I do want to share with you
18 the
challenges are cultural.
19
[Slide.]
20
If you look at the first picture, if you
21 can
see, a perfect team, right, so we wanted to
22 work
together, so we did want to talk to each
23
other, it is important, and that is the message I
24
really want to hone in, because the challenges
25
right now we are facing, especially in companies,
82
1 is
this challenge.
2
We have been able to overcome that in a
3
small way within the PAT team, but this has to
4
occur broadly, as Helen pointed out, throughout the
5
Agency.
6
[Slide.]
7
So, I think the challenges are great, and
8 we
have to build teams by dancing together, and we
9 did
dance together--that is Joe Famulare and Doug
10
Ellsworth dancing, you will never seen them dance
11
anywhere else--and working as a team on smaller
12
projects and building a team. You
can see Chris
13
Watts smiling.
14
[Slide.]
15
That led to a team process that paralleled
16 the
efforts that we put together to develop a
17
guidance. The guidance is
different, it is a very
18
different guidance, it is not a "how to" guidance,
19 it
is a guidance developed as a framework, and the
20
guidance simply outlines a framework that reflects
21
analytical chemistry, industrial pharmacy,
22
pharmaceutical engineering principles, but in an
23
integrated way.
24
What it does is it changes quite a bit of
25
things each discipline might think about. The way
83
1 I
like to say that is if you change the way you
2
look at a thing, the thing you are looking at
3
changes, so when Tom and I were discussing, we are
4
discussing as an analytical chemist and a
5 industrial
pharmacy type.
6
When we brought engineers in, we got
7
engineering aspect, so now PAT is somewhat
8
different than any of the three views of that.
9
DR. SINGPURWALLA: It is called
the
10
Heisenberg principle.
11
DR. HUSSAIN: Yes. So, this is a draft
12
guidance which we are finalizing, and Chris will
13
talk to you about that, but I do want to sort of
14
share some other thoughts.
15
[Slide.]
16
We had very successful workshops.
The
17
Arden House conferences this year and last year
18
were very successful, but they were very emotional,
19
especially the one last year was very emotional.
20
The emotions came out first as R&D versus
21
Manufacturing, because they didn't want to talk to
22
each other, and then it come out between
23
pharmacists and engineers, so the engineers came up
24 to
me saying these pharmacist types don't know what
25
they are doing, but it was necessary because it
84
1
forced soul-searching, it forced the thought
2
processes that was needed, and many companies are
3
going through that right now.
4 So, the emotions gave into a lot of
5
rational discussion at Arden House this year, IFPAC
6
meeting, ISPE meeting, PDA meetings.
Now we have
7
several proposals, in fact, I expect by the end of
8
this summer or the end of this year, you will see
9 two
complete PAT lines, two different companies,
10
from crystallization to endproduct, complete
11
automated manufacturing, so that is how fast two
12
companies have moved, and one we have approved, and
13 Bryan
will talk to you about that.
14
The first training session is complete,
15
certification process is ongoing.
We have an
16
ongoing interagency agreement with National Science
17
Foundation. We would like to
explore ways of
18
expanding this, and one opportunity that has been
19
created is a new initiative called Critical Path,
20 and
we will share that with you next time.
21
The Critical Path Initiative focuses on
22 the
need for research in three areas: to
improve
23
drug development itself. One of
those is
24
industrialization, that is where PAT fits in, and
25 we
want to use that as a means to sort of highlight
85
1 the
need for public funding for research,
2
especially academic research in this area, and hope
3 to
do so in the next several months and years.
4
We had an ongoing CRADA with Pfizer on
5
chemical imaging. Things are looking
good there,
6 and
we hope to bring some of the results back to
7 you
for some sharing of that with you.
8
We have ongoing communication and
9
cooperation with other regulatory agencies. Now,
10 our
European colleagues have formed a PAT team very
11
much like ours. They are actually
going to meet
12 the
end of this month, and they have invited us to
13
participate.
14
Health Canada has met with us and they are
15
very eager to sort of join our training session
16
next year, the next training session that we start.
17
MHLW, the Japanese are looking at it very
18
intently and things are happening on the
19
harmonization front with our trying to harmonize.
20 [Slide.]
21
Now, standards development, it was very
22
important that we have a venue to develop standards
23
that bring in the multifaceted structure, engineers
24
have to talk to pharmacists, have to talk to
25
analytical chemists.
86
1
The way we thought that will happen is
2
through ASTM, because ASTM has a lot of knowhow
3
already, so we formed a committee called E55,
4 Pharmaceutical Applications of PAT. Ali Afnan will
5
talk to you about that.
6
There is growing external collaboration
7 and
emerging support structure. ISPE and PDA
are
8
interested in PAT and are actually developing
9 programs to cover a lot of the training needs
for
10 the
next several years, we have PAT Group in the
11
AAPS, discussion group.
12
We are looking at possible collaboration
13
between AAPS and ISPE to bring the material science
14 and
the engineers together to really focus on
15
processing, strong support from IFPAC and the
16
formation of an association for manufacturers. I
17
think they are struggling with some identity
18
crisis. They call it IFPACma, so I
suggested they
19
should call it IFPATma.
20
I think this association will be helpful
21
because it will house all the manufacturers of the
22
sensors, the software, and so forth, and give them
23 a
voice, a common voice to move forward.
24
When you have an association especially
25
with a nonprofit association, we can partner with
87
1
them more easily. AICHE has an extensive
2 discussion,
and we are building on the vision 20/20
3 of
AICHE especially in processing to see how that
4 can
be leveraged.
5
A growing number of academic programs that
6
focus on PAT. Several PAT
companies and training
7 opportunities
have emerged. Pharmacopeias are
8
interested in PAT. Hopefully,
they resolve the
9
acceptance criteria first.
10
PAT is now a part of the 21st Century
11
Initiative and FDA's Strategic Plan, so I think
12
that small crystal is starting to crystallize the
13
system.
14
[Slide.]
15
The next step is guidance finalization.
16 We
are moving towards a quality system for the PAT
17
process. FDA will participate in
the ASTM.
18
This afternoon, we will discuss
19
application of PAT to the Office of Biotechnology
20
Products. I want to sort of make
sure I say this
21 in
a way that emphasizes the structure.
22
Expand the scope of the guidance to
23
include Office of Biotechnology Products. Since
24
they were not part of the training and
25
certification program, the guidance is not
88
1
applicable to them.
2
The guidance is a framework guidance.
It
3
applies to any manufacturing, whether it's biotech,
4
whether it's automobile, whether it's anything, the
5
concepts apply to any manufacturing, so it will
6
apply to Office of Biotechnology Products.
7
The reason that office is not within the
8
scope is they were not trained and certified on
9
this aspect. So, the question to
you would be how
10
would we develop a training program that will meet
11
their needs, and as we go to the second training
12
program, that will have a more biotech focus and
13
then that becomes part of the PAT process.
14
I will stop my presentation and invite
15
Chris to continue. I think in the
next two to
16
three years, we want a sunset PAT.
What I mean by
17
"sunset PAT," is that becomes a regular part of our
18 CMC
and GMP program, so it will merge with the rest
19 of
the system.
20
Is two to three years the right time?
I
21
think we will see, but the intention is that this
22 is
no longer a unique program, it is part of the
23
current system.
24
With that, I will stop. If you
have any
25
questions, I will be glad to answer, or we could
89
1
answer after Chris and others have talked.
2 Finalizing PAT Guidance
3 Training and Certification
4
DR. WATTS: Thank you, Ajaz, and
thank the
5
committee for giving me just a few minutes of your
6
time to go over what we have done in terms of
7
training and certification and moving toward
8
finalizing the draft guidance that we put out back
9 in
September of 03.
10
[Slide.]
11
I just want to take a step back really
12
quickly and just summarize some of the discussions
13
that took place at this committee and the PAT
14
Subcommittee in terms of defining what PAT is, and
15
that will really give some background on the intent
16 of
the training program and what the focus was for
17 the
training program.
18
The definition that came from this and
19
subsequently made its way into the guidance was PAT
20 is
a system for designing, analyzing, and
21
controlling manufacturing through timely
22
measurements of critical quality and performance
23
attributes of raw and in-process materials and
24
processes.
25 So, it is not just focused on any one
90
1
analytical technique, it is not focused on
2
endproduct only, it is the entire manufacturing
3
process.
4
When you think about PAT, process
5
analytical technology, that term "analytical" more
6
should be thought of as analytical thinking, not
7
just simply analytical chemistry, so we made a
8
point of emphasizing that analytical, when you
9 think about that term, you should include
not only
10
chemical, but also physical, microbiological,
11
mathematical, and risk analysis, all those
12
conducted in an integrated manner to come up with a
13
framework for controlling the manufacturing
14
process.
15
[Slide.]
16
So, with that definition, the unmistakable
17
focus of PAT is to really understand the
18
manufacturing process. What we
outlined was a
19
process is considered well understood when, number
20
one, all critical sources of variability are
21
identified and explained; number two, the
22
variability is managed by the process, and,
23
finally, product quality attributes can be
24
accurately and reliably predicted.
25
So, with that focus on process
91
1
understanding, it brings in the concept of really
2
risk management, so we consider that the level of
3
process understanding is inversely proportional to
4 the
risk of producing a poor quality product.
5
So, a well understood process then offers
6
less restrictive regulatory approaches to manage
7
change to different approaches to validation.
8
So, if you focus on process understanding,
9 we
can facilitate risk-managed regulatory decisions
10 and
innovation, not only within the Agency, but
11
within the manufacturing arena and the
12
pharmaceutical industry in general.
13
[Slide.]
14
So, having that background, I want to now
15
talk about this framework that we developed for PAT
16
that came out in the guidance, and it was a
17
framework, as I just mentioned, for innovative
18
pharmaceutical manufacturing and quality assurance.
19
We really set forth some scientific
20
principles, some basic principles and concepts, and
21
described some PAT tools that would support
22
innovation.
23 In my opinion, one of the most
important
24
aspects was the regulatory strategy that would
25
accommodate innovation, and that the primary focus
92
1
there was on the PAT team approach again which Ajaz
2
mentioned briefly, the team approach to review and
3
inspection.
4
Along those lines, we developed a joint
5
training and certification program, so I want to
6
talk to you now about that training and
7
certification program.
8
[Slide.]
9
You have already seen a few slide from
10
Ajaz on the team building aspect, really getting to
11
know one another very well, and again that included
12
people from the Center for Drugs, both reviewers
13 and
compliance officers, the field investigators
14
from the Office of Regulatory Affairs, and, of
15
course, the compliance officers and reviewers from
16 the
Center of Veterinary Medicine.
17
During this training program, it was
18
important that all 15 individuals who were part of
19
that initial training program, we went through
20
everything together, every didactic session we went
21 as
a team, every practicum we went as a team.
22
The team building obviously, everyone was
23
involved there, so there it would really break down
24 the
communication barriers, which is really going
25 to
be key to ensuring that science-based,
93
1
risk-based or risk-managed approach to review and
2
inspection.
3
A brief outline of the training program
4
that we had. Two didactic
sessions, both of those
5 were conducted here at the FDA, and three
practica,
6
again, at the University of Washington, the Center
7 for
Process Analytical Chemistry; Purdue
8
University, Center for Pharmaceutical Process
9
Research, and the University of Tennessee, the
10
Measurement and Control Engineering Center.
11
[Slide.]
12
In summary, the first didactic that we had
13 was
really just to provide a general overview of
14
some of the pharmaceutical processes, the
15
scientific basis for some of those processes, why
16
they may be necessary, to really give the team a
17
feel for what some of those unit operations
18
specifically may be trying to do to the material
19 and
what are some approaches for trying to control
20
that process.
21
Of course, there was some extensive
22
discussion on some of that process analytical
23
techniques, multivariate analysis, an in-depth
24
discussion on the background of where some of the
25
multivariate analysis techniques came from,
94
1
principal component analysis, partial e-squares,
2 how
those can be used in terms of developing a
3
control system for the manufacturing processes, and
4
then finally, a general introduction to true
5
process control from a process control engineer.
6
After that, we went to the University of
7
Washington in Seattle, The Center for Process
8
Analytical Chemistry, and the focus there was
9
really on sensor technology and development. I
10
think CPAC did a wonderful job of tying that in,
11
giving some other industrial examples, and tying
12
that into how some of these sensors may be applied
13 to
the pharmaceutical industry.
14
[Slide.]
15
To maintain continuity with the practicum
16
visits, we took some of those, the sensor
17
technology, some of the sensors that were being
18 utilized
at CPAC, and put them in the use onto some
19
pharmaceutical processes at Purdue University.
20
There, we really focused on some of the
21
experiments that we conducted were blending, for
22
example, compression, granulation, traditional
23
solids processes, how some techniques were emerging
24
that may be able to allow us to control those
25
processes on line, really understand the impact of
95
1
those processes on the final product quality and
2 how
they relate, not just to consider them
3
independently, but how they relate to the final
4
product quality as a whole.
5
After having done our experiments at the
6
second practicum at Purdue, we then took some data
7 on
the granulation process. Then, when we
went to
8 the
Measurement and Control Engineering Center at
9 the
University of Tennessee, we actually analyzed
10
that data.
11
Paul Kemperlein, who is part
of MCEC,
12
really walked us through, you know, what are some
13 of
the techniques that you maybe use, what are some
14
limitations of these multivariate techniques that
15 you
may be want to be keeping in mind when you are
16
going through the review of these applications.
17
[Slide.]
18
Finally, the last didactic, we tried to
19 tie
everything together again. We broke up
into
20
teams, developed some case studies, so that we
21
could really apply what we had learned throughout
22 the
training program, and discussed those as teams,
23 a
true team approach, a reviewer, compliance
24
officer and investigator, and really began to
25
discuss what some of the relevant issues were in
96
1
terms of managing the review and inspection
2
processes.
3
That really ended the initial training
4
portion, but by no means did we think it is
5
complete. I think continuing
education is going to
6 be
vital to the success of this team, which Ajaz
7
mentioned is really going to drive the success of
8 PAT
within the Agency.
9
Along those lines, we have monthly video
10
conferences with the people that are here in
11
Rockville and the investigators that are in the
12
field, and we try to discuss some of the relevant
13
issues that are coming out, for example, some
14
recent publications or some inspections, review
15
issues that may have surfaced, and discussed those
16 as
a team, not individually as reviewers or not
17
inspection issues individually as inspectors, but
18 as
a team.
19
We also have developed a seminar series to
20
discuss some publications that may be relevant to
21
what we are trying to do within the PAT initiative,
22
and, of course, we are using the Intranet to
23
communicate some of these publications and discuss
24
those on line, really, an easy way of communicating
25
with the entire team.
97
1
[Slide.]
2
In summary, we have, in terms of the
3
training and certification, we have completed the
4
initial training program. We are
now in the
5
process of conducting some lessons learned in terms
6 of
what we have accomplished with this, maybe some
7
additional aspects that need to be considered, and
8
some of those will be discussed with this committee
9
this afternoon in terms of expanding the scope of
10 PAT
to include biotech products.
11
Again, continuing education and
12
involvement in the next training, I think is going
13 to
be critical for this group, so that we maintain
14
links, not only with the team that we currently
15
have, but the team that we intend to build.
16
We can take some of the experience of
17
those reviewers and investigators who have
18
processed and will be processing some applications
19 and
who have gone on inspections and really share
20
those with the new group that is coming in and the
21
group that we currently have, so that we can
22
understand maybe what is the best approach for us
23 to
go in terms of taking a team to do an
24
inspection.
25
Maybe we don't need to have all three
98
1
people, maybe one or two should be sufficient, and
2 we
can do discussions over the telephone or
3
videoing to handle some issue.
4
Of course, we have involved the entire
5
team in finalizing the guidance.
In my opinion, I
6
think it was very important to get a real feel for
7 how
the reviewers felt about the guidance, how the
8
compliance officers and how the investigators felt
9
about the policy that was emerging in the guidance,
10
really how that framework was going to be
11
implemented because they are going to be the ones
12 who
are really driving things.
13
They are going to be the ones who are
14
enforcing the policy, not really enforcing the
15
policy, but making sure that the process works as
16 it
should, so that it is a least burdensome
17
approach to the industry.
18
Within the Office of Testing and Research,
19 you
heard Helen mention Dr. Khan is coming on
20
board, I think it is going to be important to
21
maintain a link to the Office of Testing and
22
Research, so that we can support policy development
23 and
future training if we develop some in-house
24
expertise and what are some critical issues that we
25 may
want to be able to focus on in terms of review
99
1 and
inspection and some of the technologies that
2 may
be developed, if we can develop some of that
3
expertise in-house, we can not only bring some of
4 the
training in-house, but also have some consults,
5 we
have expertise within the Agency that we can
6
consult on a given basis.
7
[Slide.]
8
So, building on a little bit of the
9
guidance finalization, we involved the entire team
10 in
the development of the guidance, and, of course,
11
they are going to be involved in finalizing the
12
guidance.
13
The guidance was issued in September of
14 03,
and the public comment period extended through
15
November 4th, and those comments are available on
16 the
docket. You can see all, I think there
were
17
some two dozen companies or individuals that
18
submitted comments to the guidance, and we are in
19 the
process of going through those and discussing
20
those and addressing each one of those.
21
We have included the entire team and we
22
have broken the teams down into reviewers again,
23
compliance officers, and investigators, and have
24
those address each of those and see which comments
25
they may think are most relevant and convey that
100
1
back to the policy team, so that we can move
2
forward in finalizing the guidance.
3
With that, I am going to conclude this
4
portion right here. Again, I
think we may have
5
time for some questions afterwards, and I want to
6
turn it over to my colleague, Ali Afnan, who will
7
discuss the standards development process for PAT.
8 Standards Development
9
DR. AFNAN: Thank you very much
for giving
10 me
the opportunity to be here.
11
[Slide.]
12
I am going to be very quick. The
outline
13 of
the talk is why we went with ASTM, what is ASTM,
14
what is the history of the committee, where are we
15
going with it, and I will give you some background
16
also as to how, what Chris has just said, links
17
into this process.
18
[Slide.]
19
Having focused on the processing, going
20
away from product testing, which Chris very
21
beautifully put out as PAT being process
22
understanding, we had to come up with new standards
23 and
new ways of assessing whether a process was
24
right or wrong.
25
If the process was working well, then, the
101
1
product would be right, so for that reason, we
2
began to look at alternatives to the current
3
specifications we were working with because
4
effectively, we needed standards, not
5
specifications.
6
We needed a process which included all the
7
interested parties and allowed them to come in for
8 a
balanced discussion, definition of balanced
9
discussion being that we would each have one vote,
10 it
would have a due process, and, of course, there
11 was
the NTTAA Act, the National Technology Transfer
12
Act, which mandates federal departments and
13
agencies to use voluntary consensus standards in
14
place of government standards wherever possible.
15
So, having looked at all of those, we
16
decided to look at ASTM, which had already been in
17
dialog with our other departments in the agency.
18
[Slide.]
19
So, ASTM, which now they call themselves
20
ASTM International, is an ANSI-accredited standards
21
development organization with more than 100 years
22 of
experience in standard development.
23
They actually generate standards, best
24
practices, and guides, three different things, but
25
they are all done through a peer review process.
102
1
Their offices are in West Conshohocken, and they
2
meet regularly. There is a
committee which goes
3
around to various places. This
year it is in Salt
4
Lake City, and next year it is somewhere in Europe.
5
[Slide.]
6
The history of developing the committee
7 was
that through the winter and spring of 2003, FDA
8 met
with ASTM re: development of a new committee
9 for
Process Analytical Technology.
10
In October of 2003, there was a meeting at
11
ASTM, and then in December, the first
12
organizational meeting was held at which interested
13
parties from academia and industry were present.
14
In January, the nomination and election of
15
committee officers took place.
Again, if you are
16
interested in the procedures and the processes of
17
elections or how ASTM functions, the best place to
18
look at is ASTM.org, World Wide Web.
19
In February of this year, we had the first
20
meeting of ASTM E55 Committee, and the next one is
21 in
Salt Lake City, 18th through 20th of May.
22
[Slide.]
23 What is the scope of E55? E55 pretty much
24
reflects the FDA PAT draft guidance, but the scope
25 of
the committee is that the scope of the committee
103
1
shall be development of standardized nomenclature
2 and
definitions of terms, recommended practices,
3
guides, test methods, specifications, and
4
performance standards for pharmaceutical
5
application of process analytical technology.
6
The committee will encourage
research in
7
this field and sponsor symposia, workshops and
8
publications to facilitate the development of such
9
standards. The committee will
promote liaison with
10
other ASTM committees and other organizations with
11
mutual interests.
12
What was quite interesting was it took
13
about an afternoon to come up with that, and,
14
really, we thank the industry for taking a very
15
active role in coming up with that scope.
16
[Slide.]
17
Currently, E55 has three subcommittees.
18 One
is E55.01, which is PAT Systems Management;
19
E55.02, which is Systems Implementation and
20
Practice. The Executive
Subcommittee is 90, and
21 then
there is a third one, which is E55.91
22
Terminology.
23
[Slide.]
24
The Chair and the elected officers, which
25 was
by ballot effectively, of E55, the Chairman is
104
1 Don
Marlowe from the Office of the Commissioner.
2 The
Vice Chair is Ray Scherzer from GSK. The
3
Membership Secretary is James Drennen from Duquesne
4
University, and the Recording Secretary is Gawayne
5
Mahboubian-Jones from Optimal Industrial
6
Automation, Ltd., a system integration company.
7
[Slide.]
8
The Subcommittee officers.
E55.01's chair
9 is
Ken Leiper, Vice Chair is Gerry, the Secretary
10 is
Chris Watts. E55.02 Chair is Ferdinando
Aspesi
11
from Aventis. The Vice Chair,
from AstraZeneca, is
12 Bob
Chisholm. I am the Secretary.
13
E55.91, which is the Terminology
14
Subcommittee, has Larry Hecker, Abbott, as Chair,
15 and
Jim Fox, of GSK, as its Secretary.
16
There are also 8 members at large, who
17
serve on the E55 Main Executive Committee, and they
18 are
appointed from industry and academia.
19
Thank you.
20 Rapid Microbial Methods
21
DR. RILEY: What I would like to
do this
22
morning is give you a brief update on the status of
23
rapid microbiology methods as part of the PAT
24
initiative.
25
[Slide.]
105
1
As you may know, rapid microbiology
2
methods were not originally part of the PAT
3
initiative. We were sort of
looking at rapid micro
4
methods in a parallel track with the development of
5 the
PAT initiative, but finally, someone recognized
6 it
would make sense to have rapid micro methods as
7
part of PAT, so at the October 2002 PAT
8
Subcommittee meeting, there was an extensive
9
breakout session dealing with rapid microbiological
10
methods.
11
A number of speakers discussed the
12
importance of rapid microbiology methods, how they
13
could fit into PAT and also the best way to look at
14
rapid microbiological methods for the
15
pharmaceutical industry.
16
[Slide.]
17
From that point on, we worked to try to
18
integrate rapid microbiological methods into the
19 PAT
initiative because PAT had sort of a headstart
20 on
us. So, the first thing we did was
looking at a
21
training session for rapid micro.
To do that, in
22
July of 2003, here in Rockville, we had a training
23
session.
24
We invited people from CDER, ORA, CBER,
25 and
CVM to attend. As an agenda, we had an
106
1
overview of rapid microbiological method
2
technologies, a very extensive overview.
We had
3 two
rapid micro method vendors come in and talk
4 about their products and how they can be used.
5
We also had a company come in and talk
6
about their experiences of validating a rapid
7
microbiological method for pharmaceutical use.
8
[Slide.]
9
Since the team approach is very important
10 for
PAT, one of the things we had to do was to form
11 a
rapid micro method team for PAT. That
team
12
consists of Bob Coleman, expert drug investigator
13
from ORA; Dennis Guilfoyle, a pharmaceutical
14
microbiologist from the North East Regional
15
Laboratory at FDA, Brenda Uratani, a microbiologist
16
from the Office of Compliance, CDER, and myself.
17
[Slide.]
18
As we were doing the training and setting
19 up
the team, we were also in contact with a large
20
global pharmaceutical manufacturer who was
21
interested in using a rapid microbiology method for
22
their pharmaceutical manufacturing process.
23
We had a number of meetings with them to
24
discuss their use of these rapid micro methods, how
25
they would validate them, how they would submit the
107
1
information to the Agency, that sort of thing, and
2
these meetings culminated with a formal
3
presubmission meeting with the applicant in 2003,
4
where they discussed what they would submit and how
5
they would submit it.
6
Because what they wanted to do was to use
7
some different rapid micro methods for release
8
testing of a variety of non-sterile drug products,
9
they wanted to use these at multiple manufacturing
10
sites, it was decided that a comparability protocol
11
would probably be the best way for them to submit
12
this information to begin with.
13
A comparability protocol is simply a
14
written formal experimental protocol where, in this
15
case, what they are demonstrating is that their
16
rapid method is equivalent to or superior to the
17
traditional method they have been using, and it
18
talks also about the experiments they will do and
19
also the acceptance criteria that they would want
20 to
use to demonstrate that equivalence.
21
So, what they did after this meeting was
22
they submitted two comparability protocols, one for
23
product release testing for several non-sterile
24
drug products, and also testing for pharmaceutical
25
grade waters.
108
1
After the approval of the comparability
2
protocol for product release testing, they then
3
submitted a changes being affected supplement to
4
implement that rapid micro method for one of their
5
non-sterile drug products.
6
[Slide.]
7
It was decided as part of this application
8
process that an inspection would be done related to
9 the
rapid micro method implementation, and because
10 of
that, the rapid micro method team had several
11
meetings, one in September of 2003, where we mainly
12
discussed the comparability protocols that were
13
submitted by the company, and then finally, in
14
early February of 2004, we talked about the actual
15
inspection itself, what we would do, how we would
16 do
it, that sort of thing.
17
The inspection took place in late February
18 of
2004. It was led by again Bob Coleman
from the
19
Office of Regulatory Affairs, and Bob's experience
20 and
his leadership in this process was very, very
21
helpful to us especially on the inspection process.
22 It
made it go very smoothly.
23
We looked at the rapid micro method
24
itself, how it was validated. We
looked at just
25 the
general microbiological laboratory aspect of
109
1 the
pharmaceutical manufacturing facility, and also
2
looked at some of the GMPs related to the
3
manufacturing of the product that they would be
4
using the rapid micro method test for.
5
The inspection found no significant
6
problems. There was no 43 issue
as a result of
7 that
inspection, and we thought everything went
8
well both from our standpoint, as well as the
9
firm's standpoint.
10
[Slide.]
11
What is the future of rapid microbiology
12
methods in the pharmaceutical industry?
I think
13 the
ultimate goal, the ideal would be real-time
14
testing to provide immediate feedback.
I think
15
that would be very, very helpful.
16
Where are we today? The traditional
17
micro methods require several days to several weeks
18 to
get results. The current available rapid
micro
19
methods that are available today, and can be used
20
today, significantly shorten that time to result.
21
It can be as little as a day or maybe a
22 little
bit more than a day, and some of the rapid
23
methods can give you results in as little as a
24
couple of hours.
25
We think even though it is not real-time
110
1 testing, it still provides much better
control,
2
much better understanding of the manufacturing
3
process from a microbiological standpoint and
4
hopefully, can help detect and enable you to
5
correct a potential problem before it becomes a
6
real and serious problem as far as microbiological
7
quality of the drug product is concerned.
8
We are hoping that our experiences that we
9
have had so far with our rapid micro method
10
submission and inspection and approval process will
11
encourage others in industry to also use this PAT
12
regulatory pathway to look at other rapid micro
13
methods and use them to improve their manufacturing
14
process and understanding.
15
I thank you for your attention this
16
morning and I guess we will take questions of any
17
presentations of this session.
18
Committee Discussions and Recommendations
19
DR. MEYER: One question for Ajaz
and I
20
guess one for Chris.
21
As the U.S. develops this PAT concept and
22
begins to apply it, it seems like it is better to
23
harmonize as things are being developed than after
24
they are set in stone.
25
Is there an effort with the Japanese, the
111
1
Europeans, the Canadians to harmonize on the front
2
end?
3
DR. HUSSAIN: Yes, in terms of I
think
4
there is quite a significant dialog and discussion,
5 and
I think the framework provides a way forward
6
because as a framework, it does not get in how to,
7 and
harmonizing how-to guidance is a difficult
8
challenge, so this is the time to do this.
9
That is the reason we felt ASTM
also
10
provides a way forward because the devices, the
11
Center for Devices, for example, utilize the ASTM
12
standards, and these are international standards,
13 so
many of the members on the ASTM committees are
14
international members right now, Europe and U.S.
15
right now, and we are encouraging people from Japan
16 to
join in.
17
So, that would be a way forward, so you
18 are
absolutely correct. I mean we are trying
to do
19
that as you move along, and the progress has been
20
significant on that. That is what
I tried to say
21 is
we are harmonizing without trying to harmonize.
22
DR. MEYER: My question to Chris,
if I
23
understood you correctly, there is about a
24
15-member team, a variety of disciplines, that were
25
sent through this fairly intensive training
112
1
program?
2
DR. WATTS: Correct, yes.
3
DR. MEYER: Will that be all there
is, or
4 how
is this going to grow to be 150 people or will
5 it?
6
DR. WATTS: Well, as Ajaz
mentioned, I
7
think within a few years, two to three years, he
8 envisions
it being a regular part of the operation
9
within the CMC review and GMP inspection when it
10
comes to this team approach to PAT.
11
We have every intention of expanding the
12
training program to include more members within
13
CDER, the Office of Pharmaceutical Science, Office
14 of
New Drug Chemistry, Office of Compliance, but I
15
think the immediate need may be to expand the scope
16 to
include the Office of Biotechnology Products,
17
which will be included in the discussion this
18
afternoon.
19
Based on a lot of the comments that we got
20
from the guidance that we issued in September,
21
there were a significant number of comments
22
suggesting that we do expand the scope to include
23
OBP, and as far as an immediate need, I think that
24 may
be more urgent in terms of expanding the team
25
concept.
113
1
DR. COONEY: Another question on
the
2
education side, actually, two questions.
Could you
3
comment a bit on what do you see as the important
4
metrics that you use in measuring the success of
5 the
educational program and then could you also
6
elaborate a bit on what do you see as the major
7
challenges in continuing to evolve and develop the
8
educational program?
9
DR. WATTS: Actually, I think one
of the
10
most important aspects was just the team approach.
11 The
technical aspects will be actually rather
12
simple to address when it comes to terms of getting
13
some expertise either within academic environment
14 or
within industry that have given technical
15
expertise that can convey that to the team.
16
Given the team approach, rather than
17
expecting one member to have all the answers, then,
18 as
a team, we think we can have most of the right
19
questions, we can ask most of the right questions,
20
just not having one person have all the right
21
answers.
22
As Ali has said on many occasions, the sum
23 of
the team is much more than the individual
24
components, so it is much more than just what each
25
member brings to it.
114
1
A real metric, again, I think the team
2
approach, that was one of the most important
3
aspects, can they communicate as a team, can they
4
really work as a team, for example, with the rapid
5
micro inspection process.
6
That is relatively a novel concept when it
7
comes to the regulatory environment.
Typically,
8 the
reviewers are responsible for review only,
9
inspectors are responsible for inspection only.
10
There is little, if any, communication between the
11
two.
12
What we are really treating it as is a
13
two-way street, not just reviewers participating on
14
inspection, but what are some of the key aspects of
15 the
manufacturing process that an inspector may be
16
familiar with that they can convey to other members
17 of
the team.
18
Really, I think the communication with the
19
team is one of the most important aspects, the
20
technical aspects or the scientific aspects, which
21
will be a little simpler to address, I think, with
22
training.
23
DR. COONEY: Just one more
point. In the
24
training exercises, do you present problems of
25
innovation or scenarios where you would not expect
115
1
previously people to be able to have had all the
2
answers and then ask them to try and synthesize a
3
strategy or an approach?
4
DR. WATTS: Actually, some of the
case
5
studies that we developed are along those lines
6
exactly. During the second
didactic, it wasn't
7
just this is what one person did. This is the
8
problem, how would you as a team think about
9
solving that problem, not just regulating it, the
10
problem of solving it in general.
11
DR. KOCH: I think the question of
12
developing metrics will become increasing important
13
just in observing the first class that went
14
through, the team building indeed was there. As
15 you
go to 150, it is going to be more difficult to
16
dance, there is going to be more variation.
17
The first group was exceptional.
If every
18 one
of the 150 projected fits that description,
19
it's a wonderful program. I think
I have to add,
20
too, the team building exercise that you went
21
through before the training, that was I think
22
replaced by a team building that occurred, say, if
23 I
look at the practicum and the didactic, it was
24
quite obvious that the team members were very
25
conscious to make sure that everybody on the team
116
1
understood the technology to a working level, and
2 it
wasn't as if two or three came away with
3
understanding it and didn't bring the others up.
4
It was very obvious that by the end of the
5
program, they were quite excited to move ahead, and
6
that is where the problem I think in the future is
7
going to come, is that as you grow the number in
8 the
team, you have to develop more metrics to
9
evaluate how well it is going.
10
A small number is relatively easy, I
11
think, to build the teamwork especially as it is
12
getting off the ground.
13
DR. KAROL: Bryan, I would like to
ask you
14 a
little bit about the microbial methods.
That is
15
very exciting that you are moving to real-time
16
detection.
17
Can you tell us a little bit about the
18
processes that will be involved, what you are
19
thinking of, and are there particular organisms
20
that will be difficult to detect? You know, where
21 are
you having your problems in moving in this
22
direction?
23
DR. RILEY: Well, right now I
think the
24
methods that we are looking at are fairly simple
25 and
straightforward. We are not going to do
117
1
anything too exotic to begin with.
A lot of the
2
methods, even the rapid methods are still growth
3
based, they have an enrichment step, and then an
4
alternate detection method to detect fairly small
5
numbers of microorganisms.
6
But I think as we get into some of the
7
more exotic methods that don't rely on any growth
8 at
all, you know, cytometry, that type of thing, I
9
think the issue is going to be again how do you
10
measure, you know, make sure you detect everything,
11 and
look at how are we going to validate that, how
12 are
we going to make sure that that is possible.
13
DR. KAROL: I wondered if you were
moving
14
into DNA technology or any of the molecular biology
15
techniques now.
16
DR. RILEY: It is for some of the
17
identification. What I have
talked about mainly
18 has
been the enumeration or
19
qualitative/quantitative type tests, but certainly
20 for
identification, yes, a lot of people are
21
looking at that using nucleic acid methods,
22
sequencing, PCR, that sort of thing, for detection
23 or
identification of organisms, and that I think is
24
becoming much more common, and it is something that
25 I
think we are encouraging, as well.
118
1
DR. KIBBE: Anybody else?
2
DR. HUSSAIN: Why don't we finish
with the
3
committee questions before the audience?
4
DR. KIBBE: If we could hold off
for a
5
second and see if there is anybody else on the
6
committee.
7
DR. COONEY: I have a question on
the
8
rapid microbial. Do you also have
an interagency
9
cooperation with Homeland Security, in this area,
10 as
well? There seems to be a synergy.
11
DR. RILEY: We don't really have a
direct
12
formal connection at this point although one of the
13
team members has been involved in that, so I am
14
hoping that we can work something from that to get
15
more involvement in our aspect of it.
But you are
16
right, it does go together, a lot of those types of
17
rapid methods that they would be interested in are
18
things that we could apply, as well.
19
DR. KIBBE: Anybody else on the
committee?
20
[No response.]
21
DR. KIBBE: If you could come to
the
22 microphone
and identify yourself, and then let us
23
know what your question is.
24
DR. CHERNEY: Hi, I am Barry
Cherney of
25 the
FDA.
119
1
My question was essentially the same one
2 as
was just asked by the committee members, I know
3 the
CDC and other federal agencies, DARPA, are very
4
interested in the rapid microbial techniques and
5
have made actually a lot of advancement in that,
6 and
I was also wondering what we have done to get
7
involved in those type of efforts as an overall
8
approach for the Federal Government.
9
DR. RILEY; I agree. I think we are
10
starting to do that. Certainly,
within FDA, we are
11
looking at some of the different centers to see
12
what they are doing, but you are right, other
13
government agencies have done a lot of work along
14
these lines, and we need to have more of a coherent
15
approach or at least cooperation and information
16
sharing between the different agencies and
17
different groups that are doing that, and I think
18
that will be very helpful for everybody.
19
DR. KIBBE: Ajaz, you had
something to
20 say.
You leaned forward like you were poised.
21
DR. HUSSAIN: I think what would
be useful
22 is
if you could share some thoughts in terms of how
23 do
you think we have progressed so far, especially
24 Tom
and Judy, and folks who were on the
25
subcommittee, what we could have done better or
120
1
what we should we be looking out for in the future,
2
that would be very helpful.
3
Also, as part of this, I think there are
4
external leverages that really have to come
5
together here, not only in the international arena,
6 but
also in terms of academia, in terms of public
7
funding for some of the research that is needed
8
especially in pharmaceutical manufacturing, and so
9
forth, how do you recommend we move forward in many
10 of
these areas.
11
DR. KOCH: I guess I would make
one
12
suggestion, and that is not to lose the momentum
13
that started with the training of the first group,
14 and
I know that the second group hasn't necessarily
15
been put together yet, and there is obviously good
16
reasons for that, but don't lose that momentum
17
because it is a growing area.
18
DR. KIBBE: We have two observers
from
19
industry, what does industry think?
20
MR. MIGLIACCIO: I guess I would
just
21
comment on the training, that I think one of the
22
frustrations that FDA has is the number of
23
applications and supplements that are coming in
24
from industry.
25
The good news is, I think Chris had a
121
1
slide that said PAT equals process understanding,
2 and
we are 100 percent behind that. What we
are
3
doing now is using, in the framework that the
4
guidance has provided, we are using PAT for process
5
understanding, and we are putting all our resources
6
into that, identifying sources of variability and
7
dealing with them, not necessarily moving to
8
primary control of our processes.
9
So, I think there is some frustration that
10
they are not seeing as many supplements.
Right now
11 you
probably have enough people trained to deal
12
with what you are getting. I
think once our
13
resources can move from process understanding and
14
process capability into primary control, then, you
15
will start seeing more supplements coming in and
16
more new drug applications coming in.
17
DR. KIBBE: Anybody else? Comment?
18
DR. BOEHLERT: I was going to make
a very
19
similar comment. You know, there
was a lot of
20
initial interest. A number of
large companies very
21
interested in the techniques involved with PAT
22
making presentations. I am wondering if that is
23
starting to wane, you know, if the FDA has seen a
24
steady influx of companies asking for information
25 or
did it start off high and then it is sort of
122
1
drifting off.
2
The other issue on the microbiology, I
3
think there is probably considerable interest on
4 the
part of companies in that technique, but there
5 is
some constraints around it right now, and those
6 are
compendia tests that are different, and I think
7
there needs to be some interaction with the
8
pharmacopeia on some of these topics because there
9 are
different endpoints.
10
Even though you can demonstrate
11
equivalency, the compendia test right now doesn't
12
cover the rapid micro technique.
13
DR. KIBBE: Do you have a response?
14
MR. MIGLIACCIO: Yes. On the is the
15
interest waning, absolutely not.
In fact, the good
16
news is if you have seen the transcripts of any of
17 the
recent industry meetings and presentations over
18 the
last year or so, we have gone from talking
19
about concepts to talking about applications, and
20
there are many more applications out there right
21 now
of PAT where people are either solving
22
20-year-old problems or looking at a new way to
23
make a new product.
24
So, it is moving forward. The
interest is
25
increasing exponentially right now.
It is a matter
123
1 of
once someone introduces in the public an
2
application, others are grabbing onto those
3
applications and bringing them home, so I think it
4 is
increasing significantly.
5
Ajaz.
6
DR. HUSSAIN: I totally agree with
that,
7 and
I think what we have seen is I think the
8
requests we get for presentations have skyrocketed,
9 so
we cannot handle most of it, so we are actually
10
refusing--not refusing--we are trying to be very
11
selective in where we speak.
12
I think others have taken up the charge
13 and
that is wonderful, and that is the reason why
14 we
feel that I think we don't have to keep speaking
15 all
the time, and we have other champions that have
16
been created, and the champions are coming from
17
industry, academia, and everywhere.
18
The number of questions being asked of FDA
19 is
increasing, and the number of proposals that
20
people are coming forward with is increasing. So,
21
right now, for example, we do not have many, we
22
have seven or eight proposals right now, which will
23
translate into some very focused comparability
24
protocols and other aspects, so at least seven or
25
eight by the end of this year.
124
1
DR. KIBBE: Tom.
2
DR. LAYLOFF: First of all, I
think that
3 the
number of people trained is probably more than
4
appropriate for the amount of material coming in.
5
I think the industry is under an
6
imperative to move to just-in-time manufacturing
7
because of the model that Wal-Mart has put out, of
8
essentially maintaining zero inventory at their
9
level, which means that the inventory control has
10 to
shift back to the producer, which means that
11
they have to be able to bring things more to
12
just-in-time, and PAT is going to be able to handle
13
that or make it better anyhow, reduce the dwell
14
time, which is going to be critical for maintaining
15
good supply and keeping inventory costs down.
16
I think the initiative has gone very well
17 so
far. It has to hatch on its own case, on
its
18 own
time, otherwise, the momentum will fall apart.
19 So,
I think as the industry moves, and you move
20
with it, it will develop and expand.
21
DR. KIBBE: Introduce yourself.
22
DR. RITCHIE: Gary Ritchie. I am with the
23 USP
and currently the liaison with the process
24
analytical technology project team that was formed.
25
There were some questions or issues raised
125
1
directed to the compendia barriers, I suppose, and
2
what I just wanted to do with the committee was
3
just to let them know that the project team is
4
addressing some of those issues, one with respect
5 to
rapid micro methods, a second one with respect
6 to
I think the content uniformity issue, and,
7
third, I guess in general, other techniques that
8 may
be perceived currently as general chapters or
9
proposed that may be barriers, and that there is a
10
work group that will be looking at those areas, and
11
doing what we can do to see if we can improve or
12
remove those barriers.
13
I just wanted to make that comment and let
14 the
committee know that it is being actively looked
15 at.
16
DR. KIBBE: Thank you.
17
Tom, did you have something else?
18
DR. LAYLOFF: This is a comment
more on
19
compendia issues. The compendia
or market
20
standards, the part of the law, and occasionally,
21 you
run into unusual circumstances because of
22
incorporation of standards and laws, and probably
23 the
most exciting ones I have ever attended was the
24
protein equivalent to nitrogen and the analysis of
25
grain for protein equivalents is a kilodalton
126
1
determination is done and the nitrogen is
2
determined.
3
There is a number called a PETN, the
4
protein equivalent to nitrogen, the little
5
multiplier. Well, it turns out
the multiplier was
6
wrong, and it was a decision to change the number,
7 and
the number was off by 2 to 3 percent, something
8
like that.
9
It was one of the most heated meetings I
10
have ever attended because everybody said if you
11
change that number by 2 or 3 percent, you change
12 the
value of millions of tons of grain in ships and
13
barges and warehouses everywhere.
14
So, legal standards, even though they may
15 not
be correct, cannot be changed in a very
16
cavalier fashion because they involve a lot of
17
work, a lot of impact, and the same is true for the
18
USP, there are many methods that are obsolete, but
19 if
you change them immediately, all the firms that
20
have worked away from using those and validated
21
against them, are now in a box of having to
22
revalidate all their processes against the new
23
standards.
24
DR. DeLUCA: Before making my
comment, I
25
would just comment I wonder what was the basis for
127
1
that value in the first place, did it have peer
2
review.
3
With that little comment, you know, what
4 we
are talking about here, manufacturing process,
5 for
a long time, we have tried to bring science
6
into the manufacturing area, and this is certainly
7 an
opportunity to do that. I mean this
requires
8
science.
9
I think science requires scholarly work
10 and
publications, and it seems that what I have
11
heard today, an awful lot of work has gone into the
12
PAT, but I am not so sure that we have seen
13
publications coming out of this work, and I think
14
this has got to get into the literature.
15
So, I think we need to encourage that.
16
Along those lines, we are. We
recognized this I
17
guess a little over a year ago that we wanted to
18
have an actual theme issue devoted to this in Pharm
19
Sci. Tech, and Ajaz is the editor along with Tom
20
Hale of that theme issue.
21
What we are trying to get publications,
22
people who are actually doing research in this
23
area, and it seems with all the presentations that
24
have gone on, some of the conferences and whatnot,
25
that we could solicit from these people, and there
128
1 is
people around this table here who probably could
2 be
contributors to this, certainly, we would like
3 to
encourage the industry to submit their work in
4
this area.
5
So, I think this is essential to have
6
this, to get this kind of research and science into
7 the literature, the rapid microbiology
methods,
8
these would be great publications.
9
I think the important thing about it, that
10 you
would have some peer review of these, so you
11
wouldn't maybe make some mistakes about having a
12
value for the nitrogen and protein correlation if
13 you
had that kind of critique.
14
DR. KIBBE: Bryan, you had a
comment?
15
DR. RILEY: I just wanted to
respond to
16 the
question about USP and possibly not meeting USP
17
standards if you use a rapid micro method.
18
I don't think it is as big a concern as
19
some people may think it might be because even
20
though some of the rapid methods may use a totally
21
different basis of measurement and give you a very
22
different number than the traditional USP microbial
23
limits test or whatever, I think that you can
24
certain compare, when you are assessing the
25
usability of a rapid method, you can compare it to
129
1 the
results you are getting with the USP method and
2
certainly set your acceptance criteria based on the
3
fact that you are looking at different numbers, and
4 that even though a product can still meet
your
5
acceptance criteria with a rapid method, it would
6
still meet the acceptance criteria if you use the
7 USP
method even though the numbers may be very
8
different.
9
So, I think that should be taken into
10
account and compared when you are assessing the
11
method itself.
12
DR. BOEHLERT: I agree, I think
the issue
13 is
around equivalent to or better, which is how USP
14
defines alternate tests.
15
DR. RILEY: Yes, and I think
demonstrating
16
equivalence to the USP test should not be that
17
difficult for a lot of the rapid methods.
18
DR. LAYLOFF: With regard to the
testing
19 for
viable organisms, the rapid tests will
20
frequently give false positives.
Do they also give
21
false negatives?
22
DR. RILEY: It can depend on the
test and
23
what you are testing. It is
something that has to
24 be
looked at on a case-by-case basis, if you are
25
looking at a product or you are looking at water,
130
1 you
could have interference, that sort of thing.
2 It
really depends on what you are looking at.
3
As I said, there are some growth-based
4
rapid methods, and those would have very
5
similar--if you are looking at growth in the media
6 or
not, that is going to be very similar to the
7
growth-based traditional compendia test.
8
Some of the rapid methods that don't
9
require growth, it looks like a viable stain, that
10
type of thing, that is something that we would have
11 to
determine experimentally.
12
DR. LAYLOFF: But that would be a
false
13
positive rather than a false negative, or do you
14 get
false negatives also?
15
DR. RILEY: I think it depends on
the
16
method.
17
DR. SINGPURWALLA: You wanted to
answer
18 two
questions, are we on the right track and any
19
recommendations. Well, I just
need a point of
20
clarification. It has much to do
with I don't
21
understand what PAT is all about.
22
So, the first question to you is how is it
23
different from process control practiced in
24
automobile industries and manufacturing industries,
25 and
if it is the same, I am surprised that the drug
131
1
industry has not been using it because my sense is
2
that the drug industry has been using it ever since
3 I
was a student.
4
MR. MIGLIACCIO: What has happened
over
5 the
last five, seven years is we have the
6
analytical technology, so the near infrared has
7
been there, and statistical process control has
8
been there.
9
What has been absent is the engineering
10
solution to bring the technology right to the shop
11
floor to marry the analytical technology to the
12
manufacturing equipment. That is
what we have now
13 in
process analytical technology.
14
So, you are doing real-time process,
15
monitoring, and control versus taking samples,
16
bring them through a laboratory, and then doing SPC
17 on
that.
18
So, there is a paradigm shift that we have
19
gone through, that you have real-time monitoring,
20 and
not just of a unit dose sample that you have
21
taken out of a blender or 10 tablets that you have
22
taken off a tablet press, but of a very large N.
23 The
N has increased substantially our ability to
24
monitor the process.
25
DR. HUSSAIN: I think that is a
good
132
1
point. At the same time, I think the key aspect has
2
been that in the sense some have regarded that the
3
pharmaceuticals would be quite different, I mean if
4 you
really look at some of the literature, the
5
thought process had been that pharmaceutical dosage
6
forms are different from making machines, and so
7
forth, so some of those principles might not apply.
8
So, it has been an evolution, it has been
9 a
paradigm shift, and in many ways, I have used the
10
phrase testing to document quality to quality by
11
design. We have always talked
about quality by
12
design, but our mentality has been testing to
13
document quality, because that is what we could do.
14 I think the pharmacopeial structure,
the
15
regulatory structure had sort of reinforced that
16
thought process on that, and Gerry is right in
17
terms of when you bring the analytical tools, the
18
engineers, everybody together, it is a paradigm
19
shift, and it is happening now to a large degree.
20
DR. SINGPURWALLA: So, am I
correct in
21
understanding that you are using what the engineers
22
called "control theory" techniques into the
23
pharmaceutical industry, which was not there early
24 on?
25
DR. HUSSAIN: I think "not
there" is not
133
1
probably the correct characterization in the sense
2
different segments have different levels of
3
controls, for example, manufacture of the drug
4
substance material API, which is more closer to
5
chemical synthesis, chemical industry, you have a
6 lot
more of that in there.
7
Biotechnology evolved later on, so they
8
have more of that already in place, because process
9 is
so critical. So, there are segments, the
10
pharmaceutical dosage forms, you know, tablets,
11
capsules, and so forth, have not received the same
12
level of attention, and that is new for these
13
dosage forms, so it depends on which part of
14
industry you look at.
15
DR. SINGPURWALLA: So, to come
back to
16
your original thing, about your question, so when
17 you
say PAT, this is a generic thing.
18
DR. HUSSAIN: Yes.
19
DR. SINGPURWALLA: Not specific to
the
20
drug industry.
21
DR. HUSSAIN: Well, the framework
is
22
generic to manufacturing irrespective of which
23
manufacturing. The language, the
vocabulary we
24
have used in the guidance is pertaining to the
25
pharmaceutical industry, and from that perspective,
134
1 it
is somewhat focused on the pharmaceutical
2
situation or scenario.
3
DR. KIBBE: Anybody else? You are doing
4 so
well.
5
In light of the fact that we have run out
6 of
steam, what I propose we do is break for lunch.
7 We
have already checked, I hope we have checked,
8
with our open hearing individuals, and we are going
9 to
try to start the open to the public at 12:30
10
instead of at 1 o'clock, so that you are all
11
invited to be back here at 12:30.
12
[Whereupon, at 11:15 a.m., the proceedings
13
were recessed, to be resumed at 12:30 p.m.]
135
1
A F T E R N O O N P R O C E E D I
N G S
2 [12:30
p.m.]
3 Open Public Hearing
4
DR. KIBBE: We have how many
people who
5
have asked to speak? We have
four. Their
6
presentations, if they have slides, and what have
7
you, will be on the web page by tomorrow, we hope,
8 so
that those of you in the public who need copies,
9 and
what have you, can get access that way.
10
We have the Regional Director of the CMC
11
Regulatory Affairs from GlaxoSmithKline, Leo
12
Lucisano. We are really lucky to
have you here.
13
MR. LUCISANO: I don't have any
slides
14
today. Really, my comments are
intended to
15
complement Dr. Riley's presentation on rapid micro
16
methods this morning.
17
It wasn't until I saw Dr. Winkle's metrics
18
this morning that I realized that GlaxoSmithKline
19
provides about 10 percent of the annual reports
20
that is reviewed by new drug chemistry and about 5
21
percent of the manufacturing supplements.
22
So, we create a lot of work for the Office
23 of
Pharmaceutical Sciences. So, I was
delighted
24
when, on February 27th, the PATRIOT team from FDA,
25 the
PAT Review and Inspection Team, completed a
136
1
week-long inspection at our facility in Parma,
2
Italy.
3
It was led by Bob Coleman, as Dr. Riley
4
mentioned. Bob is a national
expert based in the
5
Atlanta District Office, and he was accompanied by
6
three microbiologists, one of which was Dr. Riley.
7
The inspection actually was triggered by
8 the
submission of a supplemental new drug
9 application
in which we sought approval of one of
10 the
types of applications for rapid micro methods.
11
That technology was endorsed as PAT technology by
12
this committee back in October of 2002.
13
The inspection was a success and now it
14
enables us to potentially implement this technology
15
across a global supply chain, and it represents the
16
culmination of a 16-month effort between the Office
17 of
Pharmaceutical Sciences, the Office of
18
Regulatory Affairs, and industry in addressing the
19
challenges associated with the first PAT
20
application approved as part of FDA's Quality
21
Initiative for the 21st Century.
22
What I wanted to do today was just spend a
23 few
minutes talking about the challenges that we
24 had
in working with the Agency to reach this
25
milestone. They were really of
three types. There
137
1 was
the technical challenges, the regulatory
2
management challenges, and the educational
3
challenges.
4
Just by way of background with respect to
5
microbiological testing in the industry, we use it
6 in
a lot of different ways. We may use it
to meet
7 the
regulatory specifications to release a drug
8
product, we may apply it to the testing of
9
excipients, such as water for injection prior to
10 its
use in the drug product, and we may also
11
utilize it to verify that a manufacturing suite is
12
sufficiently clean prior to the next phase of
13
production.
14
So, the conventional methods typically
15
take about four to seven days to complete and
16
oftentimes really represents the rate-limiting step
17 in
our decision process associated with batch
18
release or release of a manufacturing area.
19
So, with the availability of commercial
20
instrumentation for rapid micro methods being
21
available and providing results in a matter of
22
hours using technologies, such as ATP
23
bioluminescence and solid phase cytometry, there
24 are
tremendous opportunities for us in utilizing
25
rapid micro methods.
138
1
So, the technical challenges. One
of the
2
examples that we had was trying to apply rapid
3
micro methods to regulatory specification for a
4
class of dosage forms, so in working with Dr. Peter
5
Cooney's staff in the Office of Microbiology, we
6
agreed on an approach that adopted a two-stage
7
acceptance criteria, combining the qualitative
8
rapid micro method with the currently approved
9
microbial limit test that used more conventional
10
methods as stated in the USP, so if a batch of drug
11
product tested free of bioburden using the
12
qualitative rapid micro test, that no further
13
testing was required.
14
If the screen indicated the presence of
15
microorganisms, then, the conventional microbial
16
limit test was performed to determine compliance
17
with the regulatory specifications.
18
So, when you think about a high-volume
19
product where the historical data indicates that
20 the
product typically is free of bioburden,
21
applying rapid micro methods in this strategy
22
offers very significant advantages with respect to
23
product release and inventory turnover.
24
The foundation for the validation of this
25
methodology was actually provided by the PDA
139
1
technical report published in 2002.
This document
2
addressed the evaluation, validation, and
3 implementation
of new microbiological test methods,so,
4
speaking to Dr. DeLuca's comment this morning
5
about the availability of published literature
6
actually facilitating working through some of the
7
issues that we had around process analytical
8
technology and its approval.
9
The second type of challenge that we had
10 was
the regulatory management process. We
were
11
interested in applying rapid micro methods in a
12
variety of different ways at multiple FDA-approved
13
facilities, so this scope of interest would
14
potentially affect the entire approved product
15
portfolio expanding over 140 approved new drug
16
applications.
17
So, in the conventional regulatory review
18 and
approval process, this might require an
19
equivalent number of new drug applications, each
20
containing a data package demonstrating the
21
application of rapid micro methods to the specific
22
product of interest.
23 So, how would we progress rapid micro
24
methods without further overburdening the Office of
25 New
Drug Chemistry with additional supplemental new
140
1
drug applications?
2
Actually, the solution was provided by the
3
Agency with the issuance of the draft guidance on
4
comparability protocols that was published in
5
February of 2003. So, a
comparability protocol is
6
essentially a plan that evaluates the effect of
7
changes on an approved product.
8
You don't have to include product-specific
9
data, but describe the analytical procedures that
10 you
intend to use and the acceptance criteria that
11
will be achieved to demonstrate that there is no
12
adverse effect on product quality.
13
So, what we agreed upon that we would
14
submit a plan, a comparability protocol to apply
15
this technology, and we do it by a prior approval
16
supplement.
17
Once the Agency approved that plan, we
18
could then implement that technology at any GSK
19
manufacturing site that had a satisfactory CGMP
20
inspection status with the FDA, so that if these
21
conditions were met, any site within the
22
manufacturing network could adopt rapid micro
23
methods according to its own timeline and notify
24 the
Agency via a regulatory submission that does
25 not
require prior approval, such as an annual
141
1
report or changes being effected in zero day
2
supplement.
3
This agreement applied regardless of the
4
number of NDA-approved sites or number of
5
NDA-approved products and dosage forms manufactured
6 at
a particular facility.
7
So, the process, the end result was a
8
streamlined management process for adopting rapid
9
micro methods or really any process analytical
10
technology, for that matter, across a global supply
11
chain.
12
It offered advantages for the Agency by
13
decreasing the number of prior approval supplements
14
that needed to be reviewed, and also retained the
15
appropriate checks and balances for the Agency to
16
conduct an inspection at its discretion and verify
17
that the manufacturing site has fulfilled the
18
validation requirements approved in a comparability
19
protocol.
20
The last challenge was one of education,
21 and
as the discussions evolved between GSK and FDA,
22 we
recognized that there was a need for both the
23
Agency and GSK to educate their personnel regarding
24
rapid micro methods, their science, and their
25 regulation.
This was achieved in a number of
142
1
different ways.
2
Dr. Riley mentioned this morning about a
3
day-long seminar that the FDA conducted with a
4
large number of FDA staff, talking about
5
instrumentation, bringing in consultants,
6
discussing their application.
7
We also had a half-day technical meeting
8
between GSK scientists and FDA scientists in May of
9 last
year with the objective of that meeting to
10
share the ongoing work that was evolving with rapid
11
micro technology, but there was also a need to
12
educate the global organization within GSK since
13 the
regulatory process that was approved for rapid
14
micro methods was very different from the
15
conventional post-approval process for implementing
16
changes.
17
We also had to review our changed
18
management systems to assure that they could
19
accommodate the conditions of approval agreed upon
20
with the agency. The regulatory
management process
21
approved for the implementation of rapid micro
22
methods has implications for other process
23
analytical technologies in the future.
24
Consequently, we have functional groups
25
within my organization requesting the Regulatory
143
1
Affairs Department to educate them under rapid
2
micro experience with FDA, and to guide them with
3
respect to future PAT applications and their
4
approval.
5
Sixteen months from the time that this
6
advisory committee endorsed rapid micro methods as
7 a
process analytical technology, we now have an
8
approved regulatory process that can be applied
9
across the company's entire product line at any
10
FDA-approved facility.
11
It required the review and approval of
12
three supplemental new drug applications and an FDA
13
inspection by the PATRIOT team.
14
I would like to thank this advisory
15
committee for really providing the incentive to
16
address the technical, the regulatory management
17 and
educational challenges associated with the
18
approval and implementation of a PAT technology
19
platform.
20
The resolution of these challenges
21
required the application of new guidance documents,
22
such as the guidance on comparability protocols,
23 the
availability of published scientific
24
literature, such as PDA Report 33, and a new way of
25
conducting business that really required some
144
1
introspection and some adjustment on both the
2
Agency's part and ours.
3
I know within GSK, we are now motivated by
4
these first approvals and are looking at additional
5
applications of process analytical technologies
6
that may be more expansive in scope and require a
7
somewhat different road map, but I suspect the
8
challenges will still be somewhat the same and
9
require a similar investment of resources, cost,
10 and
flexibility to be successful.
11
Thank you.
12
DR. KIBBE: Do you have any
questions for
13 our
speaker, anybody?
14
[No response.]
15
DR. KIBBE: Well, I will hit you
with one.
16 Do
you have an estimate of what was saved in
17
man-hours or paperwork on both ends of the street,
18
like for your people and for the FDA people?
19
MR. LUCISANO: With respect to
reductions,
20 at
the FDA inspection at Parma, we shared with the
21
Agency that for one particular product, we would be
22
saving 160 kiloEuros--it was a European site--per
23
year with the application of rapid micro methods.
24
Certainly, the availability of only
25
submitting or the opportunity to submit only two
145
1
supplements rather than 140 supplements to gain
2
approval of a technology across approved product
3
line offers significant cost reductions for the
4
Regulatory Affairs Department.
5
DR. KIBBE: Anybody else?
6
[No response.]
7
DR. KIBBE: Thank you.
8
We now have two individuals from
9
Xcellerex, the founder, Parrish M. Galliher, and
10 the
Vice President, Quality and Regulatory Affairs,
11
Elizabeth Fowler.
12
MR. GALLIHER: Good
afternoon. Thank you
13 to
the advisory committee and Keith Webber in
14
particular for the invitation, and Ajaz's support
15 and
vote of confidence for our view on PAT for
16
biologics.
17
[Slide.]
18
I want to introduce Beth Fowler, who is
19 our
VP of Regulatory and Quality at Xcellerex.
20
The title of our talk is PAT for
21
Biologics, Ensuring Quality of Biologically
22
Produced Drugs.
23
I think I want to focus, what I would like
24 to
sort of conduct as sort of a brainstorm view of
25 our
opinions on PAT, to focus in on biotech drugs,
146
1
recombinant proteins produced in mammalian cell
2
systems or bacterial yeast systems, and less on the
3
traditional biologics, such as vaccines.
4
So, before I get going into what we have
5 to say, I would like to mention that PAT, to
us, is
6
much broader than the title, the words of PAT. It
7 is
not just, in our view, process analytical
8
technology that we are concerned about, it is a
9
broader vision of ensuring product quality across
10 all
stages of manufacturing, but also throughout
11 the
organization that is responsible for delivering
12 the
drug that comes from all parts of the
13
organization as the process is developed, as the
14
process is implemented, and as the product quality
15 is
assessed either in-line, at-line, or off-line.
16 We
will talk a bit about how, at Xcellerex, we are
17
trying to take this broader view as part of doing
18
business for ourselves and for our customers.
19
We are a contract manufacturing
20
organization for biotech products, so we have the
21
customers' product quality right square in our
22
view, and that drives a lot of what we do in our
23
business, and we find that PAT, in general, in the
24
broader interpretation, is really good business for
25 us
and for our customers.
147
1
I think looking back over the last 25
2 years
of the biologics manufacturing business, I
3
think in the eighties, the industry was consumed
4
with the simple task or the herculean task of
5
simply producing these products and the initial
6
ones getting to market, and I think the industry is
7
really consumed with that whole endeavor, which was
8
huge.
9
In the nineties, more and more drugs,
10
biotech drugs, came to the market.
We now have
11
approximately 30 individual proteins that have been
12
licensed, so getting to market in the nineties was
13
really where the industry was focusing.
14
However, in the last five years, we have
15
seen the importance of speed getting to the clinic
16 and
speed getting to the market as being more and
17
more of a driver in what we hear from our
18
customers, what we have seen in our own lives and
19
biotech companies, rushing drugs to the clinic and
20 to
the market, and very much our interpretation
21
today of PAT is focused, not just on quality, but
22
also affording speed without sacrifice of quality
23 or,
in fact, improving quality, and I will talk a
24 bit
more about that.
25
However, in the next decade, we see the
148
1
indications and trends in the industry impacting or
2
bringing costs to the forefront of biotechnology
3 and
manufacturing through the advent of competition
4
with a number of similar compounds in the market,
5
through the pressures of managed health care, and
6 so
we think that PAT will actually be invigorated
7 or
stimulated by cost pressure of the industry
8
coming in the next decade.
9
[Slide.]
10
In general, that was our review of PAT,
11
again a broader vision than simply process
12
analytical technology. We are
going to talk about
13 the
importance of PAT specifically for biologics
14
manufacturing and drill down into the real value
15 and
how we are, in several places throughout the
16
organization, in our business, implementing PAT
17
principles at various levels. I
will give you
18
specific examples of those, and then wrap up with
19 some
conclusions.
20
[Slide.]
21
We view PAT as process knowledge gained
22
through process analytics and statistically
23
designed process optimization studies to begin
24
with. So, in our view, it really
begins with
25
understanding the process that is intended to
149
1
deliver a product of a certain quality.
2
So, we view PAT more as product quality
3
knowledge rather than process analytical
4
technology. The focus is really
on product quality
5
wherever it is being produced in the manufacturing
6
process.
7
Again, to begin with, we start in the
8
development laboratory by studying the parameters
9
that affect product quality and yield in a
10
statistically oriented fashion in robustness
11
studies, and I will get into that a little bit.
12
So, the processes are really designed to
13
maintain product quality or to, in fact, improve
14
product quality, and we make real efforts there,
15 and
I will give you some examples.
16
We see the advent of continual monitoring
17 to,
in fact, further ensure process control to
18
produce a product of a defined quality, and the
19
reason that we think continual monitoring is a
20
benefit is, in fact, that if there are process
21
perturbations during a lengthy step, we can analyze
22
those perturbations more quickly and determine
23
whether or not that product is in jeopardy or
24
whether, in fact, we should invest in further
25
processing to carry it to final form.
150
1
With regard to then the manufacturing risk
2 of
further processing a batch that could be in
3
danger, again, PAT, with the database that is
4
generated through the efforts of PAT, will help us
5 not
only assess the risk to product quality, but
6
also from a manufacturing economic side, is it
7
worthwhile investing in a batch that has some sort
8 of
perturbation in this processing step.
9
So, it is not just risk to product
10
quality, but in point of fact, from a
11 manufacturer's
standpoint, we are also concerned
12
about are we delivering what the customer wants,
13 are
we delivering on the contract to produce a
14
certain amount of product of a certain defined
15
quality.
16
[Slide.]
17
Just to talk a little bit about some of
18 the
key issues that we see in this business, first
19 of
all, just stepping back a bit, there is
20
biological variation in production of the material
21
that we are interested in depending on the system
22
with which you are producing the product.
23
If you are in a transgenic system, there
24 can
be animal to animal variation, and in cell
25
culture based processes, whether they be mammalian,
151
1
bacterial, or yeast or fungi, there is variation in
2 the
cell culture step.
3
It is not a surprise, the organisms are
4
very complex, they have a huge number of genes
5 whose function can impact the manufacture of
the
6
product, so we expect that in biological systems,
7
there will be inherently more variation that could
8
affect product quality.
9
There can also be unknown pathogens
10
associated with biological systems, and these, of
11
course, are an issue with regard to biological
12
safety of the product, and there can be, in fact,
13
unrelated impurities to the drug with unknown
14
activities that can, in fact, be produced by
15
biological systems at low quantities that may not
16 be
measured.
17
So, in general, we see biologics as a
18
highly variable environment within which to make a
19
product, and taking this account, it is to me quite
20
remarkable and wonderful that there are so many
21
approved products on the market today helping so
22
many people who are, in fact, in need.
23
So, we see this background therefore
24
advocating the greater value then of more process
25
analytical technology or more
product quality
152
1
knowledge given the inherent variability. So, that
2 is
the general background in which we see the value
3 of
PAT.
4
[Slide.]
5
To just talk a bit more about product risk
6
management, this is our present state of affairs.
7
First of all, in today's biologics
8
manufacturing, we generally start with a viewpoint
9
that minimal process change should be attempted or
10
even allowed as the drug moves through the clinic
11 or
into the clinic and on to the market, we try to
12
minimize the number of process changes.
13
Two. Process parameters are used,
that
14 is,
process control parameters are generally used
15 as
surrogates for product quality indication or
16
monitoring. That is, we are not
really directly
17
monitoring the product quality attributes in every
18
step of the manufacturing process mainly due to
19
limitation in analytical technology and specificity
20
especially in the presence of crude background
21
materials and matrices that interfere with current
22
methodologies.
23
Therefore, we rely on post-production
24
release and some in-process testing again through
25
surrogate markers to ensure product consistency.
153
1
Again post-facto is the point, post-production is
2 the
operative here.
3
Generally, we are faced with processing a
4
batch to completion, which can be an investment of
5
millions of dollars, and then to find out that we
6
have somewhere along the way lost the product
7
quality attributes that we intended to achieve, and
8
that batch no longer can be considered releasable.
9
So, today's business is post-production
10 and
there is a huge investment made in the intent
11 of
that batch being released, but, in fact, our
12
methodologies are very large, inadequate to ensure
13
that on-line.
14
The future vision that we have is that
15
real-time, on-line or at-line monitoring of product
16
quality can provide increased assurance of process
17 in
that product consistency, and that is the vision
18
that we are very interested in.
19
We think it reduces our risk, we think it
20
improves the product quality all along the way. We
21
think the customer ultimately wants to know that
22
anyway, as we do, and at the end of the day, if you
23 add
up the dollars, it is cost efficient, and I
24
will give some examples.
25
Secondly, increased process understanding
154
1
enables risk-adjusted evaluation of process and
2
product data, so that means when we do have a
3
spurious event in manufacturing, which we will
4
have, which everyone does have, and don't believe
5
them if they tell you they are not having them, we
6 can
use the product quality analytical methodology
7
on-line to assess the product quality impact at
8
that moment and decide whether or not that batch
9
should be processed or not in order to achieve a
10
certain product quality attribute.
11
[Slide.]
12
So, let's talk about where on-line, or
13
in-line, or at-line technology stands as of today.
14
On this slide on your left, the three
15
major stages of manufacturing include fermentation,
16
purification, and formulation fill finish.
17
In the second column, the purpose of each
18
step certainly is to control product quality in the
19
fermentation and to assure biosafety, that is, the
20
adventitious agents that may impact the
21
fermentation step, purification, again control
22
product quality, impurity removal, ensure
23 biosafety,
virus clearance, bioburden clearance, et
24
cetera, and finally, in formulation fill finish,
25
ensure product quality, uniformity, and again
155
1
safety.
2 Present day, in the middle column, the
3
third column over, in the fermentation step, we
4
measure cell growth and cell viability, and we
5
measure a number of metabolic parameters that, in
6
part, control whether we use as control of the
7
growth and the viability of the cells with the
8
intent then, and the prevalidated, prospective
9
validated purpose of producing the product of a
10
certain quantity and a certain quality, but we do
11 not
measure the product quality directly in the
12
cell culture step.
13
In purification, again we measure process
14
parameters including those listed as surrogate
15
markers of control of product quality.
In order to
16
measure product quality, we have to take samples
17
off-line, purify the product, and measure its
18
quality attributes.
19
Finally, in formulation fill finish, we
20 get
a chance to really look at the product itself
21 and
the environment, the quality in terms of
22
adventitious contamination and volume as an
23
example, fill volume.
24
So, it is not really until we get to the
25 end
of the process that we really get a look at the
156
1
product quality attributes that we are trying to
2 get
at.
3
Our view for the future then is that in
4 the
cell growth culture step, in the fermentation
5
step, we want to be able to measure on-line,
6
in-line, non-invasively, because we are trying to
7
protect the fermentation from contamination, the
8
content, tbe product concentration in the fermenter
9 and
the quality. In fact, it could be a very
10
sensitive attribute of quality.
It could be
11
tertiary or quaternary structure.
It could be
12
potency, it could be glycosylation.
13
We want to understand the impurity profile
14 and
any other adventitious agents that have entered
15 the
step. I want to remind you that the
background
16 in
this step is very dirty, relatively speaking,
17
that is, there are many nutrients added to the
18
fermentation to promote cell growth. These are
19
obviously components that we need to purify away,
20 so
there is a complex chemical background against
21
which we are asking to measure the product quality
22 and
content.
23
Similarly, downstream in purification, we
24
want to watch quality all the time, the
25
concentration as well, as we are clearing the
157
1
impurities from the product and clearing any
2
adventitious agents.
3
Lastly, formulation fill finish, I think
4
what we really want to do is as we are formulating
5 our
pre-formulation, really make sure we do have
6 the
right quality of the product at that point,
7
because at that point in the process, the value of
8 the
product is very high, there has been a lot
9
invested in it, and we want to make sure that we
10 are
going to go forward, do a fill with active,
11
appropriately folded, biologically active product,
12 if
that is the attribute of the product at that
13
stage.
14
So, that gives you a vision of the future
15 of
what we are trying to achieve, and we feel there
16 are
real values to achieving that.
17
[Slide.]
18
Simply stated, we want to ensure that
19
product quality remains consistent throughout the
20
process from the beginning to the end, not just
21
measuring it after the fact.
22
We want to assess deviations and their
23
impact in real-time, which do occur.
Some of them
24 are
trivial, some of them are major, nevertheless,
25 the
cost invested in a cell culture step is huge.
158
1
It's about nearly 50 percent of the total
2
manufacturing cost is incurred within the cell
3
culture production step.
4
So, we want to avoid the cost of
5
processing unreleasable batches at that stage. So,
6
this is really cost avoidance, rapid cost
7
avoidance, not just ensuring product quality, we
8
want to kill bad batches fast and early.
9
If we want to continue processing, we have
10 got
the data set to justify the batch processing
11 and
ultimately, we will be ensured of batch
12
release.
13
Three. Continual process
monitoring
14
obviates need for process validation.
I think that
15 may
be a little bit broad claim, but I think the
16
implication of processing legal technology with
17
regard to its impact on potential, the reductions
18 in
process validation is huge.
19
It is huge to us because I can tell you
20
today, in order to validate a process, and I am not
21
just talking about the three qualification lots at
22
scale prior to BLA, I am talking about all the
23
process robustness studies and the assay validation
24
that is done to support those process robustness
25
studies.
159
1
The industry spends 50 to 100 man-years in
2
studying the impact of process perturbations on
3
product quality and process yield.
That is a huge
4
investment at the cost of a quarter of a million
5
dollars per person year. You do
the math, it's a
6
gigantic investment.
7
In addition, we validate the assays that
8
support the measurement of the product quality
9
under those conditions. So, there
is a huge
10
investment in doing process validation.
If we can
11
supplant that by doing real-time process quality
12
assessment and reduce process validation effort,
13
that is a very big implication for the industry.
14
We can reduce testing requirements at the
15 end
of the process potentially if we are measuring
16
product quality and content all along the way. I
17
know that makes perfect sense to me as long as
18
those assays are validated.
19
Ultimately, we can increase process
20
knowledge through identification of critical steps
21 and
parameters that impact quality, and this helps
22
obviously improve the risk assessment and validity
23 on
any particular batch that is in question.
24
[Slide.]
25
The investment risk is substantial.
Let's
160
1
just take on-line bioburden as an example in the
2
cell culture step. The
assumptions here are
3
listed. If we are making 20
batches a year, a $20
4
million annual budget, fully loaded,
that is $1
5
million per batch, fully loaded, a 90 percent
6
overall success rate facility, that means 18
7
batches a year gets released,
that means 2 do not,
8 so
the cost of lost batches is $2 million a year.
9
If we had on-line bioburden in the
10
fermenter that could detect the contamination
11
in-line, or at-line, or on-line at the time that it
12
occurred, we wouldn't invest in the processing of
13
that batch downstream. That could
save half the
14
manufacturing costs of a batch, that is, the
15
downstream costs, and we would go on with the next,
16
dump that batch, and restart with the next batch.
17
That is good business. It's good
business
18 for
the customer, it's good business for Xcellerex.
19
[Slide.]
20
We are involved in process analytical
21
technology activities at Xcellerex.
It turns out
22 we
licensed the technology platform that positioned
23 the
company to be in this frame of thinking in the
24 way
of doing business, and I have listed here sort
25 of
four or five main bullets that we are pursuing
161
1 at
this time under process development.
2
We are using high throughput screening to
3
statistically optimize process parameters.
4
Two. We are using process
analytics to
5
look at glycosylation, for instance, and microarray
6
technology, process control via noninvasive sensors
7
which we have developed, including pH and DO2, and
8 we
are using on-line environmental monitoring of
9
non-viable particulates in our manufacturing steps
10 in
our modular systems.
11
The benefit from optimization of process
12
development, on the right, is certainly to optimize
13 the
process from the start, to examine more
14
parameters in less time.
15
So, we are doing very large statistically
16
designed experiments now to screen many more
17
parameters that could affect product quality or
18
yield. We are using automation
robotics to do
19
that.
20
In process analytics with these real-time
21
assays, we can assess product quality in complex
22
backgrounds, not in-line, but at-line at this
23
point, and the non-invasive nature of the sensors
24
allows us to avoid contamination of the process
25
stream.
162
1
Ultimately, on-line will bring us
2
real-time assessment of environmental parameters
3 and
control, as I mentioned.
4
[Slide.]
5
So, specifically, in manufacturing--sorry,
6
that was more of a focus on process development
7
analytics--but in the manufacturing, what we have
8
really implemented in automation include electronic
9
batch records, non-invasive sensors, and on-line
10
quality assurance.
11
So, again, this is showing us a broader
12
view of PAT, so at manufacturing level, we are not
13
just talking about on-line sensors and on-line
14
activity, on-line analytics, we are talking about
15 an
overall quality attribute or quality program
16
that achieves what we think is a higher level of
17
product quality, and automation is one vehicle to
18 do
that.
19
We are using controlled environment
20
modules to separate the operators in the process,
21 and
we are using disposables. On the right
is
22
listed the benefits of doing business this way.
23
So, this is our sort of approach to
24
process analytical technology, but again thinking
25 on
the broader level of product quality knowledge
163
1 and
improvement in the manufacturing.
2
[Slide.]
3
With regard to knowledge management, and
4
data, trending, and archiving, we have put in a
5
system that is getting right at that, and so we are
6
right in line with the PAT
philosophy of using
7
process knowledge historically, archiving it,
8
trending it, statistically analyzing it with our
9
eFactory platform.
10
[Slide.]
11
In process optimization, here is another
12
example. We do many multivariate studies with many,
13
many combinations of variables.
This is a graph of
14 one
experiment in which we have cross-plotted the
15
results of duplicates in one experiment in which we
16
have looked at over 300 different media
17
formulations.
18
Through these methodologies, which give us
19 the
statistical data and power, shown in the
20
numbers on the right, lower right, we really get a
21
good look at process robustness and parameters that
22
affect product quality and yield.
23
So, the automation and robotics puts us
24
ahead in terms of understanding more about our
25
process before it goes into manufacturing.
164
1
[Slide.]
2
Electronic batch records. Here is
a
3
picture. This gives us the
ability to provide
4
on-line quality assurance, which is again advocated
5 by
the PAT guidance. We use electronic
batch
6
records to catch compliance issues with the
7
operators, signatures, quantities, process control
8
parameters, so that real-time, we are catching
9
product or process control parameters that are out
10 of
spec or out of control.
11
[Slide.]
12
We look at our data historically.
Here is
13 a chart of 30 batches or so, and there are
four
14
lines on the graph showing the data from different
15
parameters that we are measuring.
In fact, this is
16 a
composite graph of the step yields of the
17
process, so there are four steps in each graph of
18
data points, is showing the step yield for that
19
particular step in the process.
20
The point here is that we are using
21
statistical process control as advocated in the PAT
22
guidance to learn about the process, to learn more
23
about what affects product quality, product
24
performance, and product yield.
25
[Slide.]
165
1
A couple of more slides. I would
like to
2
just mention the challenges in applying process
3
analytical technology to biologics.
Really, I
4
think are three or four key
points.
5
First, the investment in bringing
6
analytics on-line is not trivial.
We don't see a
7 big
driver to do that, and I think that, as I
8
mentioned, cost drivers will, in fact, I believe
9
stimulate more investment in on-line or at-line
10
technology.
11
I think until we have cost pressure in the
12
industry, there will not be a huge driver to do
13
that.
14
Secondly, innovation to develop analytical
15
tools to assess critical attributes really is where
16 it
has to start. It has to start back in
the lab
17 long before you get to the manufacturing
line, you
18
have got to be back in the lab converting the
19
technology to something that is applicable on-line,
20
perfecting that technology, miniaturizing it, and
21
designing it to work in the plant floor.
22
Extensive data has to be accumulated then
23 in
order to validate the methodology to identify
24
those critical attributes and appropriate limits
25 for
that on-line method.
166
1
Always, there is a regulatory uncertainty.
2 I
think we are always concerned about more data
3
revealing more variation, and why would we want
4
that if, in fact, the variation is out of spec.
5 So,
that is always a concern, and I think it is a
6
matter of a lot of the date.
7
Stringency of limits related to the
8
criticality of impact gets to how widely you are
9
going to validate the variance tolerance in your
10
process with your on-line methodologies.
11
[Slide.]
12
Again, the regulatory risk is data, how
13
much data is too much data, what is the collection
14
interval, continuous versus intermittent data
15
collection. How to use that data,
speeding release
16 or
speeding off-line release post-batch, or
17
real-time release, and how to manage noise.
18
At the end of the day, we do not want to
19
lose product itself, we do not want to lose
20
productivity or lower the plant output.
It just
21
leads to higher manufacturing costs and lost
22
product quality for the client.
23
[Slide.]
24
Here is an example of a continuous
25
real-time data set with spurious spikes.
This
167
1
process actually tested the in-control, but, in
2
fact, we had spikes during the continuous
3
monitoring, are those spikes meaningful.
4 The organization needs to have a
mechanism
5 by
which to analyze spurious spikes due to
6
electronic noise or other things in order to ensure
7
that it is an issue or not an issue.
8
[Slide.]
9
In summary, we think that the impact of
10 PAT
is as follows. First, we clearly want to
11
measure the product quality in the process stream,
12 and
we support that.
13
Secondly, we want to increase the
14
understanding of the process and the product
15
quality relationship. There is a
relationship. It
16 is
not just product by process, or process by
17
product. The two go together and
understanding
18
more about that is money in the bank.
19
Third, continual process monitoring
20
obviates the need for process validation. We think
21
that is possible within limits.
22
Fourth, we believe that PAT enables
23
science-based decisionmaking real-time in
24
manufacturing where it has a huge value.
It can
25
reduce batch release time ultimately at the end of
168
1 the
batch, and can ultimately increase plant
2
capacity. It can overall lower manufacturing risk
3
and, in fact, lower our cost of goods delivered.
4
So, in summary, we think after all those
5
things, PAT technology really can be a very cost
6
effective investment for a manufacturing
7
organization.
8 Thank you.
9
DR. KIBBE: Thank you. Are there any
10
quick questions? Then, we will
move on to our next
11
speaker.
12
DR. COONEY: Parrish, do you see
any
13
particular needs in the guidance that have been put
14
forward so far on PAT to extend it to biologics?
15
MR. GALLIHER: Yes, we had a
discussion
16
actually a week or so ago with Ajaz and team. I
17
think the impact of reduction in process validation
18 is
understated in the guidance as written.
We
19
would like to explore that further and perhaps
20
expand the interpretation and the understanding of
21 the
impact biologics.
22
The cost of process robustness study and
23 the
cost of validation is huge, and it may or may
24 not
be the right way to go ultimately if we are
25
really thinking about PAT. So, I
think that is
169
1
particularly an area for biologics that I would
2
think about.
3
DR. SINGPURWALLA: I have two
comments.
4 One
is your control charts. It looks like
your
5
step one is out of control, right?
6
MR. GALLIHER: That data does not
show any
7
release parameters.
8
DR. SINGPURWALLA: The top one, to
me it
9
seems like it is out of control, but that is a
10
minor point. The major point is
this, that we have
11 had
two talks, one by yourself, one by the previous
12
speaker, and what we have seen is extolling the
13
virtues of PAT into your particular industry.
14
It is my sense that the FDA has taken the
15
initiative and the lead in terms of infusing PAT
16
into the pharmaceutical industry.
It has made you
17
more efficient, presumably you saved some money.
18 How
much of that money has trickled down to the
19
consumer as a consequence, or is there any estimate
20 of
that? Because it is government
investment in
21 the
end.
22
MR. GALLIHER: I am not sure I can
answer
23
that directly. I would say that
in biotech
24
manufacturing, cost pressure is not really present,
25 so
consumer cost reduction interest in the
170
1
pharmaceutical, at the end of the day, has not
2
really trickled back to manufacturing organizations
3 as
part of biotech pharma companies saying to
4
manufacturing you must lower costs.
5
The game has been to get to
market quickly
6 or
to the clinic and to produce enough product.
We
7
have not seen on a broad scale yet the trickle-down
8 of
high cost of drugs, biopharmaceutical drugs to
9 the
manufacturing floor. It has not really
10
happened.
11
That is why I said at the beginning of the
12
talk, I think that is in the next decade. As
13
managed care begins to trickle back down through
14 the
pharmaceutical value chain to the manufacturing
15
floor, we will begin to see it.
16
DR. SINGPURWALLA: So, what has
been the
17
gain then?
18
MR. GALLIHER: The gain for
manufacturing,
19 the
gain for the organization?
20
DR. SINGPURWALLA: Yes.
21
MR. GALLIHER: The gain for the
22
pharmaceutical organization is to reduce its
23
operating costs and therefore, presumably,
24
hopefully, to increase profits.
25
DR. SINGPURWALLA: Ah, but I want
to see
171
1
some of your profits come to me.
2
[Laughter.]
3
MR. GALLIHER: Well, maybe we
should have
4 a
talk outside.
5
DR. KIBBE: Anybody else?
6
DR. SELASSIE: I have a broad
question on
7
your statistical process control.
Are they
8
sequential and are those the overall yields for the
9
whole process?
10
MR. GALLIHER: This is an example
of a
11 process development data set, where in process
12
development, again, this is where we are building
13
information about the process, this is not actual
14
manufacturing runs.
15
We are looking at the performance of
16
different steps in the process and the yield. The
17
lines that go through the data points are averages
18 of
the data.
19
DR. SELASSIE: I am kind of
curious
20
because it looks like as you go from one step to
21 the
fourth step, I mean the yields gradually go
22
down. Is that the overall yield
or just the yield
23 for
each step?
24
MR. GALLIHER: Each line is the
step
25
yield. Again, this is listed here
as an example.
172
1
DR. SWADENER: Since this is a
sense of an
2
evaluation of the process, I am presuming that some
3 of
this is used to determine whether some steps are
4
necessary or not in the monitoring process?
5
MR. GALLIHER: Well, what we do is
we look
6 at
this, this is the kind of data that we look at
7 to
determine whether or not there are trends
8
starting to impact the manufacturing controls, so
9
instead of just looking at a few data points at a
10
time, we look over a number of data points to
11
determine if there is a trend developing in the
12
data.
13
We have shown this graph as a process
14
development data set illustrative of the process of
15
looking at many data points over a long period to
16
determine if there is a trend in the data, in the
17
performance of the process that you wouldn't see if
18 you
were just looking at a few batches at a time.
19
DR. SWADENER: Do you sometimes find that
20
some of your data points that you thought were good
21
data points, were not good data points, therefore,
22 you
don't need to use them?
23
MR. GALLIHER: No. I mean if we are
24
producing a pharmaceutical drug for intended human
25
use, each batch is tested and has to meet with
173
1
these criteria before it is ever released.
2
DR. SWADENER: But suppose one
data point
3
consistently comes out with the same results all
4 the
time, and doesn't tell you much?
5
MR. GALLIHER: Well, each assay is
6
validated to ensure that it is measuring the
7
intended attribute of the product, so we are quite
8
sure that that is not happening.
9
In those particular assays, there are
10
controls that are included in those analytical
11
assays to ensure that the analytical method is, in
12 fact,
valid every time it is run.
13
DR. SWADENER: What I am saying is
suppose
14 a
given data point consistently comes up with the
15
same result, and it is not really adding any new
16
data to the whole process, can you drop that one
17 and
move it somewhere else?
18
MR. GALLIHER: Not without going
through
19 the
program of change control, which is a regulated
20
method of evolving process analytical technology or
21
release assays or process methodologies or
22
controls.
23
DR. KIBBE: I think we need to
move on.
24 We
are gaining back all the time we saved this
25
morning.
174
1
Thank you very much.
2
We have a representative from the
3
pharmaceutical segment manager of Siemens Energy &
4
Automation, Troy Logan.
5
MR. LOGAN: Good afternoon. I would like
6 to
start by thanking the committee for providing
7 the
opportunity to speak here today about some of
8 the
experiences that Siemens has had with process
9
analytical technologies.
10
[Slide.]
11
The PAT opportunities that are listed in
12 the
PAT draft guidance published by the FDA are
13
that it can help to reduce production time, to have
14
faster production lead time, also right first time
15
quality, which means that the whole quality system
16 is
an integral part of the process, and a kind of
17
quality system built in by design.
18
Also, managing variability, trying to
19
reduce the variability of the process to have a
20
more consistent process.
21
Facilitating continuous processing meaning
22
that we can move faster from one unit operation to
23 the
next with fewer waiting times, which most of
24 the
time are due to laboratory tests.
25
We can increase automation to improve
175
1
operator safety and reduce human errors, which is
2
more of a risk consideration.
3
Then, the ultimate goal is real-time
4
product release. In fact, to
achieve real-time
5
product release, we need to achieve the first steps
6
listed above.
7
[Slide.]
8
Real-time product release means that we
9 can
release the product to the market without a
10
final test, so without doing laboratory tests, but
11 just
by reviewing process characteristics.
12
[Slide.]
13
If we consider the whole biopharmaceutical
14
process, there are a few steps which are very
15
important and which have a big impact on the
16
quality of the product.
17
For instance, the bioreactor stage is one
18 of
the most important steps because it has a large
19
impact on the final quality of the product,
20
compared to separation and purification where the
21
quality cannot be changed very much. We can only
22
isolate the desired product out of the
23
fermentation.
24
So, in fact, the first step where PAT
25
should be applied is in steps where the impact on
176
1
quality is the highest, and this is the bioreactor
2
stage. Later in the manufacturing
process, we see
3
that the biggest impact on final product quality is
4 in
the formulation step, so that is why formulation
5
happens to be the first one addressed for PAT for
6
drug manufacturing, also known as secondary
7
manufacturing.
8
All other areas can benefit similarly from
9
PAT. The idea is to start with
the areas where it
10
will have the greatest impact and the returns will
11 be
the greatest.
12
[Slide.]
13
To achieve real-time product release, we
14
need to bring together many disciplines, and we
15
must carefully consider the capabilities of each as
16 we
do. For instance, we have to combine
17
manufacturing execution systems together with
18
advanced control systems, with process modeling,
19
also with process development, with multivariate
20
data analysis or chemometrics, with process
21
understanding and with process analytics, all, of
22
course, inside of a regulatory framework.
23
[Slide.]
24
If we look at the whole concept, there is
25 the
process layer on the bottom and the IT
177
1
infrastructure on the top. There
are two aspects
2 of
this whole PAT concept, the control aspect on
3 one
side and the process monitoring aspect on the
4
other.
5
Looking inside the boxes, we see that the
6
control solution is built out of control modules
7 and
equipment modules, brought together to form
8
pharmaceutical modules, a batch management system,
9
and, of course, electronic batch records, which are
10 fed
into the MES or IT infrastructure.
11
On the other side are the process
12
analytics which can be applied in
two ways.
13
First, for process specification verification and
14 real-time
product release, and, second, for
15
collecting information from the process to apply an
16
iterative learning control system that will help to
17
increase our knowledge of the process on the fly as
18 the
process runs and, based on that, improve the
19
control strategy.
20
Further on top, as you gain more knowledge
21
about your process, you can begin to optimize that
22
process.
23
[Slide.]
24
We look now to a real world example, that
25
is, control of a bioreactor,
which is typically
178
1
based on monitoring pH, dissolved oxygen and
2
temperature, and apply a closed loop control
3
strategy based on the information from these
4
sensors.
5
If we now also introduce a PAT solution,
6 it
can help by providing more information about the
7
process, not just secondary parameters, but also
8
chemical composition and biological performance or
9
biological status of the process. This information
10 can
then be used as an input to the control system.
11
Separate from this, there is typically a
12
laboratory that is checking the quality of the
13
product and making decisions about holding or
14
releasing the product to the market.
15
A future strategy can be that decisions
16 are
no longer made in the laboratory, but instead,
17 the
process control system on the manufacturing
18
floor decides, based on the information obtained
19
from PAT, if product will be held or released to
20 the
market.
21
[Slide.]
22
This is an example of where we have
23
applied PAT for fast identification of
24
contaminations or a certain disturbance in a
25
process. This is from a
yeast-based fermentation
179
1
where the major threat to the process is
2
contamination introduced by microorganisms coming
3 in
through air that is sparged into the bioreactor
4 or
via the substrate before it is transferred into
5 the
bioreactor.
6 The conventional laboratory test
7
normally take 8 hours before it is known if this
8
contamination has taken place.
With this new way
9 of
applying PAT, we are able to quickly, within a
10 few
seconds, identify when there is a
11
contamination.
12
[Slide.]
13
What you see here is a representation of
14
this kind of classification. On
purpose, we have
15
contaminated the yeast fermentation with 7 of the
16
most common microorganisms that, in the case of
17
this company, caused one-third of their rejected
18
batches, so that means significant economic impact
19 in
their business.
20
We intentionally contaminated the
21
fermentation and found that we can classify and
22
identify the outcome into contaminated or not
23 contaminated
product, and this chart is the result
24 of
that experiment.
25
[Slide.]
180
1
Here is another example. An
in-situ probe
2 was
placed inside a bioreactor and is monitoring
3 the
process by collecting the spectra from the
4
beginning to the end, and a principal component
5
analysis is being applied. The
principal component
6
analysis is used to monitor process change
7
throughout the batch.
8
[Slide.]
9
What you see here is a two principal
10
component plot that represents the major changes of
11
this process. From a process
control point of
12
view, we are mainly interested in what is changing
13 in
the process, so we would like for everything
14
that is staying constant to be taken out of what is
15
being monitored. That is exactly
what a principal
16
component analysis does.
17
The result of this principal component
18
analysis is a plot that is called a process
19
fingerprint. It represents a
typical batch track.
20 The
next step is to define the ideal track, which
21 is
the so-called "golden" track.
By following this
22 track,
the required endproduct quality can be
23
achieved.
24
The next step is to determine the maximum
25
acceptable tolerance to achieve the required
181
1
endproduct quality. That is this
tunnel, which can
2 be
calculated based on good batches. This
is a
3
kind of standard deviation that we will allow
4
around the process track.
5
This is a great tool for helping to define
6 if
the process is running consistently.
Also, if
7
sudden process disturbances occur, it is a fast
8
detection tool that helps to avoid lasting impact
9 of
those disturbances.
10
Shown here in the middle of this chart is
11 where we had a disturbance due to an oxygen
12
depletion when an oxygen valve was blocked.
13
[Slide.]
14
Now we have the PAT road map
15
implementation stages for the implementation of
16
PAT. It consists of three major
steps. First, is
17 the
measuring part including monitoring and process
18
understanding. The second one is
the controls, and
19 the
third is optimization.
20
Along with these three steps we have some
21
parallel tracks. One is knowledge
and change
22
management, another is the validation aspect, and
23 the
third, the people and organizational issues.
24
Because the introduction of this PAT solution will
25
cause a lot of changes in the organization, people
182
1
have to make decisions differently.
2
For example, the decision on holding or
3
releasing product will be made on the manufacturing
4
floor and no longer in the laboratory.
This means
5
that the work processes of the organization must be
6
realigned.
7
We start first with risk assessment on
8
product quality and on the process, so to determine
9 the
required product quality and assess the process
10 to
determine which process parameters are the
11
relevant ones to track.
12
The third part is the analyzer assessment,
13
finding out which analyzer is most appropriate for
14 the
type of process and what information is needed
15
from that analyzer. Once all of that information is
16
collected, a multivariate data analysis is
17
conducted. This focuses on
finding the
18
relationships between product quality and process
19 parameters.
20
Based on that, you can then begin the
21
design of experiments. The PAT
solution will then
22
begin to help to determine which are the good
23
batches and isolate the "golden" batch.
24
The next step is then control.
Here, we
25 can
modify control parameters if the process goes
183
1 off
track, and we can get an understanding of the
2
ideal process control strategy.
When the process
3 is
running off track, other techniques can be used
4 to
get the process under control again, and we can
5
improve process knowledge with the application of
6 an
iterative learning control strategy.
7
During all these different steps, we are
8
collecting a lot of information - process behavior,
9
process capabilities, process quality, et cetera.
10
This data can then be used to further
11
optimize the process meaning we can further
12
optimize the "golden" process track, perhaps the
13
processing time can be shortened, improving
14
efficiency of equipment utilization, or the process
15 can
be optimized to use fewer resources and still
16
achieve the required final endproduct quality.
17
In conclusion, the use of these PAT
18
technologies will become part of an ongoing
19
strategy of continuous process improvement.
20
Thank you for your attention.
21
DR. KIBBE: Quick questions,
anyone?
22
DR. SINGPURWALLA: Your second
slide said
23
something about production release of
24
pharmaceuticals without final tests.
25
MR. LOGAN: Yes.
184
1
DR. SINGPURWALLA: Are you serious
about
2
that?
3
MR. LOGAN: I will have to begin
by
4
answering that I am really here as a spokesperson
5 for
our technical people, but as I understand it,
6
that is the ultimate end goal that they are
7
attempting to achieve, and they are seriously
8
pursuing it with the end users that we are trying
9 to
work with.
10
DR. SINGPURWALLA: Maybe that
needs a
11
point of clarification. You don't
want to test,
12 you
cannot test every product because if you tested
13 it,
you couldn't sell it. When I buy a pill,
it is
14
presumably not tested, but then
you still want to
15
sample even though you use PAT techniques at the
16
end, you do want to sample.
17
Here is an analogy. Suppose you
are
18
building an airplane engine, it has got many parts.
19 You
test each part. There is no guarantee that when
20 you
put it all together, the engine will function.
21 So,
you still need to do testing at the end to make
22
sure that nothing has been overlooked.
No? Why is
23
that?
24
DR. KIBBE: I will allow your
colleague to
25
respond. We will let you off the
hook.
185
1
Anybody else have anything?
2
[No response.]
3
DR. KIBBE: Thank you very much.
4
Our last speaker during the open public
5
hearing is someone from Laboratory Instrumentation
6
Scientist, Foss-NIRSystems, Robert Mattes.
7
MR. MATTES: Thank you. I am Robert
8
Mattes.
9
I would like to talk to you today about
10
near infrared spectroscopy as possibly one of these
11
analytic tools that would help in the toolbox for
12
PAT, and by demonstrating some of our experiences
13 in
PAT so far as we have implemented some
14
techniques in the tableting arena.
15
[Slide.]
16
The near infrared, just so everybody
17
knows, is the region between the visible and the
18
mid-IR, and it looks at overtones of the
19
fundamental absorptions in the mid-IR.
20
[Slide.]
21
One of the things that we
have done for
22
years using near infrared has been the inspection
23 of
incoming raw materials, and we can measure them
24 for
identification and qualification of those
25
materials, so that you are making sure that you
186
1
have the right materials going into a process
2
fermentation cell before starting a reaction.
3
One of the things we also have had a lot
4 of
experience in and has been implemented in
5
manufacturing environments is measurement of
6
moisture content and lyophilized product. I am
7
going to show some data from each of those.
8
[Slide.]
9
I have a similar chart to the last speaker
10
here that shows, first of all, the typical types of
11
monitoring that we do real time In a process
12
reactor for temperature, pH, oxygen level, you
13
know, and you are controlling the temperature and
14
sparging and pH level.
15
With the near-infrared probe also
16
introduced directly into the process reactor, we
17 are
able to measure analytes, amino acids, glucose
18
levels, feedstock levels in that process reactor
19 real
time, which helps the manufacturing people a
20
great deal.
21
We haven't actually installed this in, in
22
research laboratories and plants so far.
23
So, if you are looking at the raw
24
materials that we have now identified and qualified
25
being brought into the bioreactor, then, we are
187
1
monitoring the analytes and can real-time adjust pH
2 or
nutrient levels, and so forth, according to the
3
data that we get.
4
As the product comes out, we can measure
5 the
moisture content as the product is being dried
6 and
possibly lyophilized. This can also lead
to
7
control feedback for process improvement through
8
statistical process control charts, and so forth,
9 as
previous people have mentioned.
10
[Slide.]
11
Here are some of the organisms that we
12
have worked with, Escherichia coli, products like
13 you see on the list there, and the biomass
also.
14
With one spectrum that you take in the near
15
infrared, you can analyze multiple components
16
instantaneously with the same spectrum.
17
In fact, we are working on one experiment
18
right now where we are looking at 28 different
19
analytes including all the amino acids, glucose,
20
glutamate, lactate, and so forth.
Really, your
21
requirements are not limited in that sense.
22
[Slide.]
23 Here is an example of a process, the raw
24
near-infrared spectra of a process as it
25
progresses. As biomass increases
the y axis
188
1
spectra or the absorbance of the spectra increases,
2 as
you can see here.
3
[Slide.]
4
You see these two major peaks are the
5
water bands in the near infrared.
It is not
6
terribly informative in that form.
We usually take
7 the
second derivative of the spectra, which
8
enhances the resolution and enhances the peak
9
separations.
10
In the top set of spectra here, we see one
11
analyte progressing with time, and it is increasing
12 in
a downward direction because of that second
13
derivative that we have taken. In
the lower set of
14
spectra, we see where different analytes are
15
appearing within a process.
16
[Slide.]
17
The colors on the charts are backwards in
18 the
overhead here, but I have corrected them in
19
your handout, I am sorry about that, but the
20
biomass should be in red as you will see it
21
increasing with time in the process, and the
22
glycerol content was decreasing there.
23
So, you can see using those spectra, you
24 can
predict and measure the levels of different
25
analytes and trend them with time rather than
189
1
waiting for a week or more sometimes, waiting for
2 the
wet chemistry to come back on a process that
3 you
are running presently.
4
[Slide.]
5
Here are some of the types of things and
6 the
types of error in precision that we have been
7
able to develop. At-line, we are
talking about
8
using a peristaltic pump that pumps out of the
9
reactor and back in again.
In-line, we are talking
10
about actually having it pulled right in the
11
reactor, which is the type of work I have been
12
working on most recently with some of our
13
customers.
14
[Slide.]
15
One of the things that has been reported
16
recently in biotechnology and
bioengineering by a
17
group that worked at Strathclyde University in the
18 UK
was CHO cell fermentation, which is a very big
19
topic right now.
20
They used a small, 2-liter bioreactor
21
similar to the one you saw in a previous lecture
22
there, and they were monitoring glucose, glutamine,
23
lactate, and ammonia.
24
[Slide.]
25
This is more like the work that I was
190
1
doing most presently with about a 100-liter
2
bioreactor with a direct fermenter interface, using
3
standard Ingold port. We are
putting our probe
4
right into the sterile environment.
The probe can
5 be
sterilized right in the environment.
6
As I say, we simultaneously can get
7
results in less than one minute of up to 28
8
analytes. Previously, we have to
have developed
9 the
model for each one of those. The
10
time-consuming part is upfront on the analysis
11
rather than the real-time usually used in wet
12
chemistry.
13
This then can be turned into monitoring
14 and
closed-loop control, adding feed or whatever,
15
changing glucose levels real-time automatically,
16 but
certainly in the nearest future, will help the
17
people to know when the levels have changed to a
18
serious level within a reactor real-time.
19
[Slide.]
20
Here is the results of that particular
21
experiment with the CHO cells.
You can see the
22
precisions and ranges that were used in that
23
experiment with ammonia, glucose, lactate, and
24
glutamine.
25
[Slide.]
191
1
Here is an example of monitoring
2
lyophilized product. You
mentioned earlier that
3 not
every sample can be measured or would be
4
destroyed. In this case, we can,
we actually
5
non-invasively, non-destructively measure right
6
through the bottom of the lyophilized bottle, and
7 we
can predict the moisture content. So,
there is
8 a
possibility of 100 percent measurement in this
9
case.
10
You see this band, the largest band there
11 is
the water band. Again, it's the second
12
derivative, so it is increasing in downward
13
direction. The driest bottle
would be the red line
14
that is up at the top.
15
[Slide.]
16
Some of the benefits for the PAT
17
initiative in the biotech area.
It gives a
18
real-time analysis of sterile environments. You
19
don't have to constantly be taking samples out that
20
could lead to problems with sterility and asepsis
21 and
also the possibility of closed-loop feedback
22
process control.
23
It is not invasive and can add to the
24
process optimization, as people spoke of earlier,
25
waste reduction, and better understanding of your
192
1
process, so that they can create safer and
2
improved, more consistent product.
3
Thank you. Are there any
questions?
4
DR. KIBBE: Questions?
5
DR. SINGPURWALLA: This is very
6
interesting. So, this is a
non-invasive method of
7
looking at some particular unit, but then do you
8
have a template for what would be a normal unit,
9 and
how do you compare these templates?
10
Suppose you have a template which says
11
this is what the spectrum of a proper product
12
should be, and then you get a defective, how do you
13 say
this is defective?
14
MR. MATTES: What you are talking
about is
15 a
qualitative analysis before predicting the sample
16
quantitatively? Yes, we can build
libraries, and I
17
have done quite a bit of work like this recently.
18
You want to build a library of what
19
qualified good samples of spectra should look like,
20 and
if it doesn't conform to those criteria,
21
statistical criteria that you have developed in
22
your library, it gives you some sort of indication,
23 or
it will not give you a prediction as such, so
24 you
won't be predicting on the wrong type of
25
spectrum.
193
1
DR. SINGPURWALLA: What you need
is a
2
template which measures the spectrum of a good
3
product versus the spectrum of a defective product
4 and
the criteria for seeing how diverse those two
5
are, because, you know, a little diversity, you
6
cannot say it's bad or good, but you need proper
7
criteria to say that this is very diverse, and I
8
don't know if you have that, but it is interesting.
9
MR. MATTES: Yes. We use statistical
10
criteria in our library model developments, and we
11 can
use bad samples as reject sets, so we can test
12
both positive and negative sets, and it is
13
basically, to use the simplest example, if you had
14
normal distributions of 3-sigma outlier or you
15
choose some number of standard deviations from the
16
mean-centered spectrum of this acceptable
17
population.
18
DR. KIBBE: Anybody else?
19
DR. COONEY: An extension of the
previous
20
question. You are using the
sensor to measure
21
multiple components in variable and complex
22
systems. To what extent do you
have to go back and
23
redevelop the algorithm for each system for the
24
components versus being able to use standard
25
wavelengths or a template, as was asked, that you
194
1 can
apply across different processes?
2
MR. MATTES: Well, each unique
process
3
really needs the model development done for that
4
process, so this, as I say, is the upfront
5
time-consuming portion of this model development,
6 but
you simultaneously are looking at all the
7
variance caused by all the different constituents
8 or
analytes in the matrix of your fermentation.
9
So you need many samples, reference
10
samples, to help you be able to do this, because
11 you
are going to have so many degrees of freedom,
12 you
need more samples.
13
DR. KIBBE: Thank you. The table at the
14
back end is just references?
15
MR. MATTES: Yes, it is just a
16
bibliography that has some references including the
17
work that Strathclyde University did on the CHO
18
cell mammalian culture.
19
DR. KIBBE: Great. Thank you very much.
20
Now we are back to the PAT Applications
21 for
Products in the Office of Biotechnology
22
Products, and we are going to start off with Keith
23
Webber. Keith is here ready to
lead the charge.
24 PAT Applications for Products in
the
25 Office of Biotechnology
Products
195
1 Overview and Issues
2
DR. WEBBER: Good afternoon. I am Keith
3
Webber and I would like to thank the committee for
4
taking the time today to participate and listen to
5 the
issues surrounding our desire to implement PAT
6
technologies for the products in the Office of
7
Biotechnology Products.
8
As Ajaz mentioned this morning, the PAT
9
guidance specifically excluded the biotech
10
products, that are regulated in our office, from
11 its
scope.
12
To some extent, this was to expedite the
13
publication of the document and also the training
14 and
qualification program for inspectors and
15
reviewers, but as he also said, it is a technology
16
that is certainly amenable to any manufacturing
17
process, so there is not inherently any reason why
18 we
couldn't implement it with these products if we
19
have the technologies and the information and
20
understanding available.
21
[Slide.]
22
Now, this afternoon, just to give you a
23
brief overview of the agenda here, I am going to
24
give an overview basically of the biotech products
25 and
the manufacturing processes for the products
196
1
that are regulated in our office, and then Dr.
2
Joneckis from CBER will give a brief overview
3
related to some of the products that are regulated
4 in
CBER.
5
After that, two members of the committee,
6 Dr.
Cooney and Dr. Koch, will give presentations to
7
describe some of the issues, as well as some of the
8
opportunities available in the area of fermentation
9 and
biological manufacturing.
10
That will be followed by Dr. Layloff, who
11
will give a brief overview of the view in this area
12
with regard to the PAT Subcommittee which he
13
chaired when it was active.
14
Afterwards, we will put up some questions
15 to
stimulate discussion. I certainly hope
that we
16
will get a good amount of discussion from the
17
committee with regard to this exciting area of
18
manufacturing.
19
[Slide.]
20
The biological products as a class include
21 all
the products listed here, which were originally
22
regulated in CBER. There was a
reorganization back
23 in
2003 that moved the recombinant DNA-derived
24
proteins, or many of them I should say, to the
25
newly formed Office of Biotechnology Products
197
1
within the Office of Pharmaceutical Sciences in
2
CDER.
3 Essentially, those are the products
that I
4 am
going to be focusing on today. Dr.
Joneckis may
5
have comments on the other products, as well, or
6 the
recombinant DNA products that are still
7
remaining in CBER.
8
[Slide.]
9
This is in terms of sort of a review.
10
[Slide.]
11
There are essentially two aspects of
12
process analytical technologies.
One requirement
13 is
that you have to have the ability to monitor the
14
critical product characteristics that are needed
15 for
the product's function, or, if it is an
16
intermediate in manufacturing, you need to be able
17 to
know what characteristics are important for
18
being able to move it forward in manufacturing to
19 the
next step.
20
Now, alternatively, there may be
21
surrogates as opposed to direct product quality
22
attributes that one can use to make decisions.
23
This monitoring, as has been mentioned a number of
24
times here, will optimally be done on-line, but at
25
this point, I think to a large extent, many of the
198
1
monitoring is done off-line, so this is something
2 we
look for in the future.
3
Secondly, one has to be able to monitor
4 and
modulate the critical process parameters to be
5
able to guide the product quality attributes and
6
quality characteristics during the manufacturing
7
process.
8
It is probably worth mentioning two other
9
requirements that may be self-evident, but are
10
certainly not trivial, that is, that you need to
11
know the critical characteristics of the product in
12 the
first place that are important for its function
13 or
that need to be obtained to get to the next step
14 in
manufacturing.
15
You also need to know how these
16
characteristics can be modified and manipulated by
17 the
manufacturing process parameters themselves.
18
That is one area that is really dependent upon
19
industry to determine during their period of
20
product development and gaining a thorough
21
understanding of their product and their process.
22
[Slide.]
23
This is really part of a come-down version
24 of
process analytical technologies, but I think has
25
most of the important aspects with regard to the
199
1
manufacturing element itself. One
has a process,
2
unit operation, one is monitoring the process
3
characteristics or process parameters, as well as,
4 if
possible, the product characteristics during the
5
process.
6
You gather this data, evaluate it, and
7
then make decisions, so that one can adjust the
8
process to ensure that the product that is coming
9 out
of that process is going to have the
10
appropriate characteristics that are desirable.
11
[Slide.]
12
This is just a brief overview, which we
13
have seen already in one of the earlier
14
presentations, of the
various
15
biotechnology processes that are utilized. This
16 isn't
all-inclusive, but are the major ones.
17
You have fermentations, harvesting from
18 the
fermenter. You have product capture from
that
19
harvest. Concentration is usually
a step that goes
20 on
after, and may be a part of product capture.
21
There are filtrations that are done often,
22
almost always chromatography of some sort, many
23
times multiple steps. There is
formulation
24
process, and if the products are lyophilized
25
products, you then have lyophilization process at
200
1 the
end.
2
I didn't cover filling operations, but
3
those are certainly amenable to PAT, as well.
4
[Slide.]
5
Now, what are the characteristics of the
6
biotech APIs that are generally considered to be
7
critical quality attributes?
Certainly, the
8
primary amino acid sequence is critical to the
9
proper functioning of the product, however, this is
10 a
characteristic that is relatively invariant,
I
11
would say, once you get into the manufacturing
12
area, and it is established at the master cell bank
13
stage or the working cell bank stage, so it is
14 usually
not looked at on a lot-to-lot basis.
15
The secondary structure pertains to the
16
local interactions between the amino acid residues
17 to
produce a structure, such as the alpha helix,
18 the
pink you see in the front, and the beta pleated
19
sheets that you see in the back, in yellow.
20
The secondary structure is really very
21
important to the protein because these are the
22
structures that serve as the building blocks to
23
produce enzymatically active sites or the binding
24
sites for protein.
25
[Slide.]
201
1
They come together, as I mentioned, to
2
form tertiary structures. This is
illustrated here
3 in
this figure by a model of an antibody FAV
4
fragment. You can see that this
is purely beta
5
pleated sheet, and these tertiary structures form
6 to
form the binding sites of the antibody itself.
7
The next level of complexity that is
8
characteristic of some proteins is the assembly of
9
independent protein molecules into multimeric
10
quaternary structures. Such
structures assemble
11
post-translationally and they are generally held
12
together by either ionic or hydrophobic
13
interactions between the independent subunits.
14
[Slide.]
15
The last, but not least certainly, of the
16 API
characteristics that I am going to talk about
17 today
are the post-translational modifications.
18
Glycosylation is probably one of the most common
19
post-translational modifications that is of concern
20
with proteins, particularly those that are made in
21 the
eukaryotic cells.
22 It is illustrated in this figure by the
23
sugar chains that are in the center of the Fc
24
fragment of an antibody molecule.
25
Glycosylation patterns and structures are
202
1
highly variable in proteins from one product to the
2
next, and they can be significantly altered, as
3
mentioned earlier, by the fermentation conditions
4
that occur during cell growth and fermentation.
5
The other modifications that
are seen in
6
proteins include the proteolytic cleavages that can
7
either be caused by endoproteinases that chew away
8 at
one end of the molecule or exoproteinases--I am
9
sorry, endoproteinases that eat the middle--may be
10
producing the final product as a necessary activity
11 to
get the product you want, or the exoproteinases
12
which eat away at the end of the protein and could
13
produce degradation products during the
14
manufacturing process.
15
There also is often or sometimes you see
16
acylations and sulfations, and many other
17
post-translational modifications that I really
18
won't describe here.
19
[Slide.]
20
Now, leaving API on its own and looking at
21 the
product characteristics themselves, which
22
really then you get into the whole impurity profile
23 of
the product and excipients that may be
present.
24
Impurities fall into two categories, the
25
process-related impurities, which are media
203
1
components coming from the fermentation process,
2
host cell proteins that would come from the
3
expression system, and then leachates, which come
4
from columns or containers that are used to store
5 the
product during processing.
6
Then, also, you have product-related
7
impurities, which are perhaps truncations of the
8
molecules or misfolded molecules or aggregates of
9 the
product, which can occur during storage or even
10
during manufacturing.
11
[Slide.]
12
Now, I would like to discuss briefly some
13 of
the analytical methods that are used currently
14 to
look at these factors for biotech products or
15
these characteristics of biotech products.
16
As was mentioned earlier, the primary
17
structure is really something that is not looked on
18 at
a lot-to-lot basis unless in particular cases,
19 you
might have, as I mentioned, a cleavage of a
20
protein that is part of the manufacturing process.
21 In
those cases, then, one generally does look at
22 the
primary structure, not necessarily with
23
sequencing, but just to demonstrate that cleavage
24 has
occurred appropriately.
25
One area that I also would note here, for
204
1
products that are patient-specific products, for
2
example, antibodies that are used for treating
3
B-cell lymphomas where each individual patient gets
4 a
unique product. There is an area where
the
5
primary structure would certainly be critical to
6 look
at as an identity test, if nothing else, prior
7 to
giving a product to the patients.
8
[Slide.]
9
The secondary structure is somewhat more
10
difficult to evaluate, and that is because there is
11 a
limited number of direct techniques. The
ones
12
that are primarily used are circular dichroism and
13 NMR
at this point.
14
Also, another complicating factor for
15
proteins is that most proteins have multiple
16
secondary structures in them. For
antibodies, it
17 is
almost all beta-pleated sheet, but other
18
proteins, you have a mixture, so you need to have a
19
method that will be able either to distill out the
20
critical values for that protein or can look at the
21
individual secondary structures separately.
22
One other complicating factor for this
23
with regard to an in-process control, which we
24
hopefully will be able to overcome at some point,
25 is
they need relatively pure material to look at
205
1
secondary structures in a protein.
2
[Slide.]
3
Now, I grouped the tertiary and quaternary
4
structures together because they are both high
5
order structures and are amenable to a similar set
6 of
analytical tools.
7
The functional assays, such as in-vitro
8
potency assays, can directly measure the
9
therapeutic--or I shouldn't say the
10 therapeutic--but
the activity of the product
11
itself, so it is semi-looked upon as a surrogate,
12 but
actually, it is a measure usually of the direct
13
activity, but it requires, of course, the
14
product-specific reagents to do that.
15
This is also true of the immunoassays.
16 You
can get a direct picture of the structure of
17 the
protein if you have antibodies that will bind
18 to
3-dimensional epitopes that are relevant to the
19
tertiary or quaternary structure, but again you
20
need to have product-specific reagents to do that.
21
Peptide mapping is a valuable method for
22
looking at the disulfide bonds to make sure that
23
they are mapped, that they are forming
24
appropriately.
25
Size-exclusion chromatography is a
206
1
relatively insensitive method for looking at
2
tertiary structure, but in some cases, you can use
3 it
to separate monomeric from multimeric forms of
4 the
protein, so that can be a very useful
5
technique.
6
Hydrophobic-interaction chromatography is
7
actually a very good method because it looks at the
8
surface charges and surface characteristics of the
9
protein and can be used to very sensitively detect
10
either misfolded proteins or proteins that are not
11
associated with their other monomers appropriately.
12
[Slide.]
13
For post-translational modifications, this
14 is
probably the most variable characteristic of the
15
protein, as I mentioned before, and analyses of
16
these usually requires a highly purified protein
17 and
some rather sophisticated methodologies, for
18
example, enzymatic cleavage and analysis of the
19
amino-linked oligosaccharide protein, however,
20
recently, the mass spec and NMR have allowed direct
21
analysis of post-translational modifications in
22
intact proteins, which is an up and coming
23
technique.
24
Peptide mapping can also pinpoint the
25
location of the modification within the protein
207
1
sequence, which is very useful for characterization
2 of
the product.
3
Immunoassays and the functional assays can
4 be
used for more impure proteins because they are a
5
little bit more specific for your product, however,
6 the
functional assays are often not very sensitive
7 to
protein modification itself unless there is a
8
specific modification that is really critical to
9 the
activity.
10
[Slide.]
11
So, to summarize, inherent challenges that
12 we
see to implementing PAT for biotech products at
13
this point are that the biotech products are
14
generally large and complex pleiotropic molecules.
15
They are composed usually of a mixture of
16
post-translational modifications, they have
17
multiple active sites. Some of
those are
18
homologous like two binding sites antibody, or they
19 can
be heterologous where you have different active
20
sites doing different functions on the same
21
protein.
22
The activities are dependent
upon the
23
complex, folded conformations of a protein, and
24
proteins are also susceptible to multiple
25
degradative events, so you need to look at a lot of
208
1
different aspects of a protein during
2
manufacturing. As I mentioned
before, these
3
include the proteolysis, aggregation, misfolding,
4
oxidation, deamidation, just to name some of those
5
that we know of.
6
[Slide.]
7
Of course, when you are considering the
8
factors involved in protein structure or actually
9 any
product, you need to consider the purity,
10
potency, and the strength, of course, but also the
11 impact that those changes or modifications or
12
variabilities to the protein would have on the
13
pharmacokinetics, the pharmacodynamics, and the
14
immunogenicity of the product.
15
That is delving more into the area of the
16
product development stage of pharmaceutical
17
development as opposed to manufacturing itself, but
18
surely, that is one of the early bits of
19
information that one needs to have, we need to
20
gather.
21
[Slide.]
22 Now, I would like to talk briefly
about a
23 few
of the manufacturing processes that have been
24
touched on before and what the current state of
25
monitor and control are.
209
1
For fermentation processes, generally, one
2 can
monitor and control the agitation rate, the pH,
3 the
ionic strength of the media, the temperature,
4
dissolved gases, media components, and by being
5
able to monitor and control those, you can then
6
control the growth rate and the expression rate
7
usually of your product.
8
This is an area where process analytical
9
technology, we will probably see it developed
10
early, because one has that control over some of
11 the
aspects of the process.
12
As we have heard before, there are methods
13 now
available for detecting or monitoring the
14
biomass and bioburden through using rapid
15
biological methods, rapid microbiological methods
16 for
sterility testing. Generally, one
monitors the
17
product by light absorbance, for example, protein
18
concentration to A280.
19
[Slide.]
20
Moving on to chromatographic processes,
21
this is again the same format.
You can monitor and
22
control your pH of the effluent or the liquid
23
phase, ionic strength, flow rate, temperature, and
24
volume, and of value here, which isn't exactly laid
25
out, though, is that because you can control the
210
1
volume and monitor the light absorbance, one can
2
then control the composition to some extent of the
3
fractions that you collect out of that, from that
4
column.
5
That is currently being done although it
6 is
really looking just at the protein
7
concentration, one usually doesn't know except by
8
doing previous experiments, to know what is in each
9 of
the fractions that you collect.
10
[Slide.]
11
Filtration processes. This
includes both
12
dead-end filtrations for removal of bacteria and
13
viruses, as well as the ultra-filtration for
14
selectively removing lower and higher molecular
15
species from a product.
16
In most cases, one can monitor and control
17 the
temperature and flow rate, the back pressure,
18 and
the volume of the filtrate, although you
19
usually can't do all those independently because
20
they are inter-related.
21
Again, we have seen before the protein
22
concentration is monitored by light absorbance and
23 the
bioburden is, at this point, generally
24
monitored off-line, but soon could be monitored
25
on-line.
211
1
Dead-end filtration is usually a
2
flow-through process and that generally allows
3
little control over the product characteristics
4
themselves other than the removal of the material
5
which is filtered out.
6
Ultra-filtration, on the other hand, can
7 be
a much more dynamic process, and that may allow
8
more control over the composition of the product.
9 For
example, ultra-filtration is often used for
10
formulation of biotech products.
11
[Slide.]
12
It was discussed a little bit earlier, the
13
lyophilization process, and this one may be
14
currently the most close to being a process
15
analytical technology. In the
lyophilizer, you can
16
monitor and control the shelf temperature and the
17
product temperature, the chamber pressure, the
18
condenser temperature, the pressure, and time in
19 the
lyophilizer.
20
The ability to monitor and control these
21
parameters allows you to control the freezing rate
22 and
the drying rate, and the moisture content, all
23 of
which directly affect the physical quality of
24 the
final product, which really is what we are
25
shooting for in process analytical technologies.
212
1
Although you have to have a product with
2 an
acceptable composition going into the
3
lyophilizer, the physical characteristics of the
4
product that comes out will play an important role
5 in
the stability and the activity of the product
6
that goes to the patient.
7 [Slide.]
8
Finally, you will see these questions
9
again at the end of our session, but I just want to
10
introduce them now, because these are points that
11 we
would like to initiate discussion with.
12
What technologies are available
now to
13
evaluate the characteristics of protein products in
14
real time during manufacturing, or to speed things
15
along with an off-line test which is faster, is
16
valuable to know, as well.
17 What tools would allow us to
understand
18 the
manufacturing process better?
19
What processes in biological drug
20
manufacturing would benefit the most from
21
implementation of PAT?
Essentially, where are we
22 going
to get the most bang for our buck, as has
23
been said before.
24
For processes or products that do not
25
currently allow direct product quality monitoring,
213
1 what other strategies would you, as a
committee,
2
recommend for product quality control in addition
3 to
control of the in-process parameters?
4
Finally, what additional elements should
5 be
incorporated in a training and certification
6
program for reviewers and inspectors of
7
biotechnology PAT applications?
8
Thank you.
9
DR. KIBBE: Does anybody have any
10
questions? It might be a good
idea for us to go
11
ahead and get at least the next speaker through the
12
process, and I think it might be useful for the
13
committee to be able to take a break then, so I
14
don't know how that does to your continuity, but it
15
would be helpful for us.
16
DR. JONECKIS: Thank you and good
17
afternoon. I am Chris
Joneckis. I am the Senior
18
Adviser for CMC Issues in CBER, Office of the
19
Director.
20
I am just going to briefly describe CBER's
21
perspective on process analytical technologies for
22 the
biotechnology and biological products that CBER
23
currently regulates.
24
[Slide.]
25
CBER regulates a wide variety of products,
214
1 as
shown in these slides, the majority of the major
2
product classes shown here. They
include a wide
3
variety of biological and biotechnology products,
4
diagnostic and processing devices, cells, and even
5
chemical entities that are clearly derived from a
6
variety of sources and manufactured using a wide
7
variety of techniques.
8
My comments today will predominantly focus
9 on
the experience that we have gained with the more
10 traditional
biologics and some of the newer
11
recombinant products that are produced from living
12
organisms and are typically extracted and further
13
modified, purified, and, for example, fill for
14
distribution following some of the examples that
15
Keith provided in the manufacturing process.
16
For many of these products, most actually,
17
product contamination with adventitious agents from
18 a
variety of sources is of primary concern, and
19
most of these products are again aseptically
20
processed.
21
It is important to point out also that
22
there are recombinant products not just in the
23
blood derivative class for the recombinant
24
analogues that CBER regulates, but also in a
25
variety of other classes including allergenic
215
1
extracts, prophylactic and therapeutic vaccines.
2
They are also used in the manufacture of various
3
cellular therapies and in some other product
4
classes not shown here.
5
CBER's approach to technology in general
6 is
also applicable to other product classes, and
7
many of the comments on PAT that I will make today
8
will be applicable to those.
9
[Slide.]
10
Historically, CBER's approach to
11
controlling the process can clearly be summed up by
12 the
mantra, if you will, that, "The process is the
13
product."
14
There has been a long historical emphasis
15 in
understanding the product and a long emphasis on
16
understanding and controlling that manufacturing
17
process. This clearly requires, not just an
18
understanding of the process and the product, but
19 the
interaction of those two, how the process
20
results in the product.
21
The nature of many of the traditional
22
biologics influenced this approach.
Many of these
23
were complex heterogeneous products susceptible to
24 a
variety of variability produced almost
25
exclusively from living sources or living sources
216
1
themselves.
2
The complex mixtures, coupled with
3 insufficient
analytical technologies, made it very
4
difficult to detect all the active components or
5
materials, in fact, that can influence the activity
6 of
the active components.
7
This necessitated a very strict control of
8 the
manufacturing process to reproducibly result in
9 the
desired product with the appropriate safety and
10
efficacy profile.
11
Recent advances in analytical technology
12 and
enhanced manufacturing processes often result
13 in
better defined products, aiding in a greater
14
assurance of producing products with the desired
15
characteristics.
16
Manufacturing is beneficial to implement
17
these newer technologies and improved approaches to
18
better control processes and demonstrate that
19
products can be consistently manufactured. That
20 was
clearly shown in many of the recently derived
21
biotechnology and biological products.
22
[Slide.]
23
An overall approach that we have followed
24 at
CBER has been that we have always encouraged the
25
application of technologies and concepts to the
217
1
manufacturing and testing of products.
2
Again, we have lived with developing
3
technologies throughout its history, and have
4
applied those to manufacturing and testing of
5
various products. We are actively
involved in the
6
development and application of these new
7
technologies.
8
Again, historically, we have developed and
9
applied technologies appropriate to specific
10
manufacturing and testing issues.
We continue to
11 be
actively engaged in developing and applying
12
these technologies.
13
For example, the conversion of older
14
technologically-based assays, such as animal-based
15
assays and cell-based assays to newer analytical
16
methods, actively involved again in development and
17
application of proteonomics and genomic
18
technologies to issues, such as product
19
characterization and adventitious agent detection.
20
This large laboratory component assists us
21 in
maintaining our knowledge base for discussions
22 in
applying these new technologies.
23
We clearly partner with manufacturers in
24
developing and implementing new technologies and
25
concepts. As I have indicated, we have had to live
218
1
with developing technology throughout history.
2
It is through these interactions with the
3
manufacturers in both development and in
4
post-approval phases that allowed the advancement
5 and
development and introduction of new
6
technologies or appropriate manufacturing
7
processes.
8
Issues are addressed, validation issues,
9 for
example, and other types of issues about
10
understanding this new technology are addressed
11
throughout the development process, the
12
post-approval process, on review and inspection, as
13
well as in review of applications.
14
[Slide.]
15
The approach to process control that CBER
16 has
emphasized is best described as a comprehensive
17
life-cycle approach to validate this process and
18
spans the life cycle of that product.
19
This approach relies on developing an
20
understanding of the process and product. Use of
21
knowledge gained can be applied throughout the life
22
cycle and typically is.
23
In addition to CBER's perspective, this
24
comprehensive approach was largely influenced
25
through interactions with manufacturers of
219
1
biologics and biotechnology products, incorporated
2
concepts and approaches often used in manufacturing
3
industries.
4
It emphasizes identification and control
5 of
critical unit operations and process variables
6 to
product intermediates, resulting in a product
7
with acceptable quality attributes.
8
Some of the elements are shown here.
They
9 are
familiar I am sure to many of you. They
also,
10 I
should point out, overlap with many of the
11
fundamental underlying principles necessary to
12
implement many of the PAT applications.
13
[Slide.]
14
As a result, over time, there have been
15
many PAT-like applications of technology to
16
manufacturing and testing. For
example, as Keith
17 had
indicated, there are many examples of continual
18
on-line monitoring of critical process attributes
19
often with real-time feedback mechanisms that may
20 be
computer assisted.
21
Within the defined parameters from the
22
validation studies and such there is also some
23
flexible control within those parameters, so one is
24 not
necessarily fixed to certain endpoints if the
25
appropriate validation characteristics support a
220
1
range within which one can operate.
2
We have been involved with application of
3
on-line analysis of various intermediates and
4
product attributes, as well as facility systems.
5
Some examples are indicated here.
6
We have approved several years ago an
7
on-line measure of a critical physical-chemical
8
quality intermediate for a naturally-derived
9
product. We have entertained
discussions, again
10
several years ago, on measuring through a
11
non-destructive method the moisture content of
12
final filled containers. We have
approved
13
appropriate physical property for changing
14
lyophilization conditions in lyophilizers.
15
CBER regulates and reviews major facility
16
changes. We approve numerous supplements that
17 described
on-line applications of water systems
18
when conductivity measurements were substituted for
19 the
wet chemistry measurements in water systems.
20
Most importantly, we have recently
21
approved microbial methods for two applications.
22
Rapid microbial methods are very concerned
23
especially to or for our cellular products, and for
24
those products where they cannot be held or stored
25
prior to the release of the sterility testing
221
1
results, so implementing methods that have allowed
2 for
a rapid turnaround to determine whether the
3
products are sterile or not has provided great
4
increase in the assurance of the quality of that
5
product.
6
I should point out that that last method
7 is
not an on-line method, but is an off-line
8
method.
9
[Slide.]
10
There are clearly potential applications
11 for
new manufacturing and testing technologies that
12
have been discussed. Many of
those advantages for
13 PAT
have been described and are probably known much
14
better to you all than to me.
15
I think some of the best applications
16 would be if one could use those in terms of
17
defining product or intermediate quality
18
characteristics. Unfortunately,
that provides the
19
most challenge and at present, I think there are
20
some great limitations to doing that in an on-line
21
fashion.
22
Immediate applications I think may be more
23
likely in terms of drug product manufacturing,
24
measuring of more single types of process or other
25
very select quality attributes.
222
1
Some of the challenges I think that we
2
face at CBER is that we still have, in contrast to
3
some of the more purified and defined recombinant
4
products, a large amount of complex and
5
heterogeneous products. All of
the issues that
6
Keith discussed about the identity, purity, and
7
composition of these products is in many cases
8
magnified when one has a complex and heterogeneous
9
product.
10 Again, I think that leads to the
ability
11
that it may be difficult to know from a multifactor
12
analysis the heterogeneous mixture, what actually
13
that relationship is.
14
Again, manufacturing unit operations in
15
biological and biotechnology products often perform
16
multiple functions. Again, the ability to measure
17 all
important product quality characteristics in a
18
continuous mode from any of those functions, I
19
think is going to be very challenging.
20
For CBER, we have the development of new
21
products, not just products within a class, but
22
again completely new products, gene therapy,
23
therapeutic vaccines, as well as cellular products.
24
[Slide.]
25
Just in summary, I think it is still
223
1
important at CBER that we understand and emphasize,
2
understanding both the product and the process,
3 clearly
integral to the development and manufacture
4 of
biotechnology and biological products.
5
The comprehensive, life-cycle approach to
6
process validation remains integral to the
7
consistent manufacture of these products.
8
Validation is still a regulatory requirement and
9
when conducted in a comprehensive life-cycle
10
manner, has provided great assurance that the
11
process will consistently produce that desired
12
product.
13
That has been most readily seen at CBER
14
when products that were approved prior to
15
validation being a regulatory requirement,
16
validated their process. They had
potential
17
savings both from economic and public health
18
perspectives.
19
We see PAT more as an extension of the
20
existing process understanding the manufacturing
21
control paradigm. I think
clearly, PAT has
22
potential applications for biotechnology and
23
biological manufacturing processes especially if it
24 can
monitor again intermediate quality attributes
25 and
provide greater assurance of that product
224
1
quality.
2
We will continue to partner with
3
manufacturers of existing and new products to
4
facilitate implementing any type of new technology
5 and
concepts, including those that can enhance the
6
knowledge and control of the manufacturing process.
7
Thank you.
8
DR. KIBBE: Does anybody have any
9
questions for Chris? Go ahead.
10
DR. COONEY: One of the particular
11
challenges for the class of products you are
12
dealing with are viruses, viral contamination.
13
How do you see some of the issues of
14
detection and validation of viral removal being
15
advanced by PAT?
16
DR. JONECKIS: That is an
interesting
17
question. Currently, I guess, for the committee's
18 benefit,
most people do challenge or clearance
19
studies, usually small scale, representative of the
20
larger scale manufacturing process.
21
In terms of detection, again, as I
22
mentioned earlier, there are efforts underway to do
23
genomic and proteomic screenings for potential
24
contaminants within products at various appropriate
25
stages in addition to the current various levels of
225
1
safety that are provided.
2
I suppose theoretically if one could with
3 the
increased sensitivity of certain methods, one
4 may
be able to do more on-line monitoring, if you
5
would, again at early or appropriate stages to
6 actually
see if there is any type of potential
7
viral materials present.
8
One could potentially in theory, depending
9
upon how much is present, again, sensitivity of
10
your methods, actually measure on-line for the
11 various
steps, present of type C retroviral
12
particles, CHO-derived products, and things of that
13
nature.
14
Similarly, you know, it has been done for
15
measuring DNA and other types of materials when it
16 is
there in a large amount in early purification
17
steps, given the sensitivity of the assay, one can
18
measure those on-line in addition to whatever model
19
studies are done to provide additional assurance
20
that your model truly reflects what is occurring.
21
DR. KIBBE: Anyone else?
22
Seeing none, I am going to take the
23
prerogative of the Chair and declare a 15-minute
24
break, which means we should be back in our seats
25 and
ready to go at approximately 2:35.
226
1
[Break.]
2
DR. KIBBE: I have been assured by
experts
3 in
the field that Tom has all the answers in his
4
presentation, so when we get to them, we will be
5
done for the day.
6
Charles Cooney is on the podium.
7
DR. COONEY: Thank you very much.
8
I am pleased to have an opportunity to
9
share some thoughts this afternoon on the question
10
that Keith Webber put before us, and that is the
11
extension of PAT to biological processes.
12
In preparing for any talk, one obsesses
13
over a number of things, one of which is the color
14 of
your tie, of course, but another is the title of
15 the
talk. I obsessed over a complex title
and a
16
simple title, and I resolved that dilemma by having
17
both.
18
[Slide.]
19
PET for PAT? The message that I
am trying
20 to
convey in my title as a place to begin is that
21
when we think about PAT and all of its virtues and
22
aspects that have been dealt with earlier today,
23
process analytical technologies applied to
24
processes and products, it is a very important
25
fundamental concept, and it means a lot.
227
1
When you think about it in terms of the
2
process, it occurred to me that we really need to
3
think about analyzing the process, as well as
4
analyzing parts of the process and the product
5
itself.
6
So, the emphasis here is to think about
7
process evaluation tools as a component of process
8
analytical technologies, and I think that the broad
9
definition that has been used for PAT very much
10
embraces that idea.
11
[Slide.]
12
In putting together my comments for this
13
afternoon, I have identified more questions than I
14
have answers, and the reason for this is that as we
15
think about going forward with the extension of PAT
16 to
biological products, there are a number of
17
issues and questions, and I would like to try to
18 put
at least a few of these into some context.
19
The first set of questions I have
20
summarized here as Some Issues.
What are the
21
issues, what is the context as we look forward, one
22 of
which is the pipeline of new products, what will
23
that look like going forward in the next 10 to 20
24
years.
25
I think there is no doubt that it is going
228
1 to
be expansive, there is going to be increased
2
complexity in the nature of the products, and it is
3
going to be a very vibrant pipeline simply based
4
upon what we see in discovery and what we see in
5
clinical trials today.
6
If we think about the increase in the
7
number of BLAs and NDAs that will be coming through
8 for
biological products, it puts a real future
9
stress on the Agency because as we look at the
10
number of these products, they are increasing
11
exponentially, and I don't think that the number of
12
people in the FDA is increasing exponentially.
13
Just a guess, but I think it's true.
14
So, what that means is that the pressure,
15 in
order to be more efficient, and to focus on a
16
risk-based strategy and understand where and when
17 to
look, at what, is really very, very timely to be
18 in
this process right now.
19
Then, of course, there is the question of
20
follow-on biologics that are beginning to--I will
21
come back to this in a moment--but are beginning to
22
come forward, and I think are going to be an
23
increasing issue.
24
Both of these issues raise the question
25 how
do biological products respond to the physical
229
1
process changes that occur when you develop a
2
process, scale the process, move it, change its
3
location, and the like.
4
We have some understanding of this, and,
5 of
course, this is fundamental to understanding how
6
biological products, particularly the complex one,
7
respond to the complex processes used to make them.
8
Underlying all this, do we have the
9
adequate analytics to address the uncertainties
10
associated with manufacturing in this industry, and
11 I
am struck by looking at the presentations we had
12
earlier today, and, of course, they all focused on
13
where we have the analytics in place.
14
In fact, do we have the necessary
15
analytics? No, I don't think we
do.
16
Are efforts underway to develop them?
17
Well, we are going to hear in the next presentation
18
that there are some very exciting efforts that are
19
underway, and I think the future looks bright, but
20 it
is only going to come with a lot of diligence
21 and
a lot of innovation in order to measure the
22
kinds of things that we really need to be looking
23 at.
24
Then, ultimately, how do we bring this
25
together to assure robustness in design and
230
1
operation of these processes.
2
[Slide.]
3
I tried to address where we are going.
4
Keith Webber already identified a number, in fact,
5 the
previous two speakers identified the range of
6
products that are out there today and the ones that
7 are
likely to be out there tomorrow, and we can
8
expect that there are going to be a lot more
9
antibodies, replacement proteins, designer
10
proteins, vaccines, not just for therapeutic use,
11 but
for prophylactic use, cellular and gene
12
therapies are being developed quite aggressively.
13
One of the other observations I would like
14 to
make, though, when we look at the range of
15
products that are there today and that are going to
16 be
there tomorrow, is that this question of
17
follow-on biologics is on the minds of many people
18 and
we need to take stock of where we are today,
19
because we really have follow-on biologics today,
20 we
have multiple processes for the same products,
21
multiple manufacturers for human growth hormone,
22
multiple manufacturers by very diverse technologies
23 for
human insulin.
24
How we have managed them is perhaps not
25 the
same way that we wish to manage them in the
231
1
future, but these are realities today, these are
2 not
things that are looming out there for the
3
future.
4
[Slide.]
5
When we look at the processes that are
6
going to be used, we have very diverse recombinant
7 protein production processes. Why are there so
8
many? Why isn't there a single
technology that has
9
emerged?
10
The answer is very simple, not all
11
processes are suitable for all products.
12
Furthermore, the intellectual property landscape is
13
such that it dictates complexity in the processes
14
that are used simply to work your way through the
15
minefield of intellectual property that is out
16
there.
17
Is that going to get simpler as we look
18
forward? No, the processes are going to become more
19
complex, driven in part by innovation, and driven
20 in
part by the nature of the products, tissue
21
products, multicellular products, and certainly the
22
potential future for transgenic plants and animals.
23
So, as we look at the array of complex
24
processes for these complex products, I do not see
25
that landscape getting simpler. I
see it remaining
232
1
complex and as a consequence, we need to be able to
2
have the analytics in place and the ways of
3
handling the data and the ways of understanding
4
these processes that is better in the future than
5 it
is today.
6
So, this leads us to a series of
7
challenges, and I have tried to organize these
8
challenges in a way that represents where we are
9
coming from and where we are going to go.
10
There is the continuing challenge of
11
rapid, cost effective development and scale-up. We
12
need to shorten the timelines, the timelines for
13
developing the processes, and if we develop better
14
processes, that should lead to improved timelines
15 for
approval of those processes, and we need to be
16
able to have more flexibility, so that the process
17 of
development and scale-up could be a lot more
18
nimble and lean than it is today.
19
But then once we have processes in place,
20 I
think the industry has done an increasingly good
21 job
in the drug space of continuous improvement,
22 and
most recently, and we have heard examples of
23
that today, PAT is a major contributor to how that
24 is
going to go forward in the future. That
is very
25
positive.
233
1
We need to understand how to better
2
achieve continuous improvement in process change in
3 the
biological space, and that is the challenge
4
that we are focusing on in this particular session.
5
What the tools that we need to do that?
6
What are the methodologies? Where
is the
7
uncertainty, and, of course, how do we understand
8
that risk, and risk is implicit in all of this.
9
Follow-on biologics present their own
10
challenges, and then when we get into complex
11
biologicals, cellular therapies, and tissue
12
engineering, there are a wide variety of unknowns
13 and
we need to understand quickly what are the
14
parameters, what are the biomarkers, what are the
15
surrogate markers, what are the direct methods that
16 we
can apply in order to get a grasp of these
17
processes and how they will define the products
18
that we make.
19
Furthermore, as we look at these
20
challenges, there is a constant tension between the
21
safety and the economic agenda, and where is the
22
proper balance in terms of how much risk we seek to
23
minimize and how much risk we seek to embrace and
24
manage and take forward.
25
[Slide.]
234
1
Well, when you look at the broad issue of
2 the
relationship between the process and the
3
product, one has to look at what goes in and what
4
comes out. We have raw materials
and environmental
5
conditions that are variables going in.
We are
6
trying to control a number of the parameters in
7
this space.
8
Some of those parameters are suitable for
9
control in a closed loop fashion.
Again, we heard
10 a
number of examples of how that is increasingly
11
important today. A number of
those parameters we
12
don't control in a closed loop manner, but we need
13 to
control them nonetheless.
14
I think the challenge in looking at this
15
very microscopic view of a process is the
16
information flow. We know how to
do process
17
control. We are going to get
better at
18
implementing new analytics on these processes.
19
There is a long history of applying statistical
20
process control and a wide variety of other
21
methodologies of process control.
22
We are going to get better at doing that,
23 and
that is all going to be incremental.
What is
24 not
going to be incremental is the more systems
25
view of understanding to do it better.
I think
235
1
there we will have some big jumps, but where I
2
think we are doing a terrible job is on the
3
information flow.
4
The information is quite an asset, a lot
5 of
money goes into generating that information, and
6 do
we adequately understand and mine it, and the
7
answer is no, we don't. In fact,
it's a very
8
poorly utilized asset, and in some cases, the
9
reason is, well, if I don't look at it, I don't
10
have to worry about the variance in it.
That is
11 one
way to control variance.
12
Another way is to say, well, let me
13
embrace that variance, let me learn from it, let me
14
capture that information, and feed that back and
15
learn, and that is an area where I think we are
16
getting better, but, frankly, I think if I look
17
back over the past decade or two, even looking at
18
work that I have done, I think we have done a
19
pretty bad job.
20
Now, what I would like to do is to stay in
21 the
frame of raising questions rather than
22
providing answers, but I can't go through a
23
presentation like this without showing some data
24 and
without taking an example to illustrate where I
25
think there are some opportunities and some of the
236
1
kind of learning that represents work-in-progress.
2
In biological processes, one of the main
3
issues we deal with is the oxygen dilemma.
4
[Slide.]
5
We all know that in most biological
6
processes, there is a requirement for oxygen for
7
efficient growth, and in this particular case,
8 recombinant protein expression. That is a given.
9
By the way, there is some interesting data
10 to
suggest that that is not necessarily true, but
11 we
won't go there now. But that is a
general
12
methodological given, and let's assume that it's
13
true for the moment.
14
But on the negative side, there is the
15
potential that if a little bit of oxygen is good,
16 is
a lot of oxygen better, and the answer is not
17
necessarily, because there is potential for both in
18
vivo and in vitro protein oxidation of methionine,
19
cysteines, for instance, and as we scale-up and as
20 we
change the amount of oxygen, as we use enriched
21
oxygen in processes, is this going to be a hazard,
22 is
it going to be a problem?
23
We wanted to explore that, and we also
24
know that oxygen can induce stress.
Actually,
25
oxygen too high or too low can induce stress. One
237
1 of
my hobbies is high altitude mountaineering, and
2 I
decided what would it be like to operate under 35
3
percent partial pressure of oxygen.
Well, I don't
4
recommend going there on a regular basis.
5
[Slide.]
6
But when we look at processes today, we
7 are
looking at scale. Traditionally, what we
have
8
done is to do a lot of our optimization of a
9
process at a shake flask scale, 100-milliliter,
10
perhaps to a 10-liter scale, and then go to 10 or
11 100
cubic meter scale.
12
The benefits of doing research at the
13
homogeneous milliliter or liter scale is that we
14 can
make the assumption that it is almost
15
homogeneous, and the work we have done over the
16
years is to better resolve events in time, so we
17
have taken analytics, like some of the probes, and
18 so
on, that have been discussed earlier, and we
19
have learned to evolve events in time and
20
understand how the time space is critical.
21
At the fermentation scale, we might do 200
22 to
300 experiments in order to get what we think is
23 an
optimum, but we really know it is not, in order
24 to
scale to the 10 to 100 cubic meter scale, but
25 all
we do is get to a place that allows us to
238
1
economically be in the business, and then,
2
hopefully, we will be allowed to undergo continuous
3 improvement
following that.
4
What I want to suggest is that what we
5
really need to think about is how we look at this
6
process development and scale-up paradigm very
7
differently.
8
That is, if we scale down, and, for
9
instance, one approach is to use reactors that are
10 100
microliters, and they indeed are homogeneous,
11 or
somewhere in that small space, and do large
12
numbers of experiments, and not just resolve events
13 in
time, but do the kind of things that were
14
described earlier, create large experimental
15
design, so that we can now not just look at our
16
experimental space, but we can look at the
17
interdependencies between the independent variables
18 in
a much more effective way, reduce the
19
uncertainty associated with how the process
20
responds to the environment, as well as changes
21
with time, and reduce the uncertainty of scale-up,
22 and
presumably reduce the variance as we do so.
23
That is not to say that we shouldn't also,
24 at
scale, resolve events that take place in time.
25
There is going to be variance in a biological
239
1
process. We can learn a lot from
that, and that
2
allows us to manage the risk associated with these
3
processes, and that goes on, as well, but we are
4
doing a better job with that than we are simply
5
going to the large-scale experimental design.
6
[Slide.]
7
A model system that we happened to choose
8 is
alpha-1 antitrypsin. It is a human
recombinant
9
protein. It is an interesting
model because you
10
notice that methionine 358 and the one at 351, it
11
sticks up like a sore thumb and is sensitive to
12
oxygen. So, we reasoned it would
be useful as a
13
molecular probe in order to determine if oxidation
14 was
a problem.
15
This molecule also actually has 10
16
methionines, several of which are partially or
17
completely exposed, and 1 unpaired cysteine that is
18
partially exposed, but with models such as this,
19
this might be the product where its structure is
20
well known, you can begin to do microscale
21
experiments that you can then project to the larger
22
scale and ask, well, what is the effect of oxygen
23 on
the molecule.
24
[Slide.]
25
In this particular case, we observed that
240
1
there was an oxygen-dependent proteolytic cleavage,
2 and
as you look at these three lines, the green
3
line is for the expression, transient expression
4
under air. The top line is
transient expression
5
under anaerobic conditions, which turns out to be
6 not
so bad. But the bottom line is
expression
7
under pure oxygen, so there is this oxygen
8
dependency of the proteolytic cleavage.
9
[Slide.]
10
How do we resolve that? Well, one
11
approach is the very hypothesis-driven problem,
12
and, of course, when you have a problem, and a
13
complex problem, and as you can see by the photo on
14 the
left, if you don't get the ropes right, you
15
could be in serious trouble, so you have got to
16
know where the problem is if you want to be in the
17
position on the right.
18
[Slide.]
19
So, how do we resolve that? Well,
we have
20 a
hypothesis. In this case, we tried many
21
hypotheses. I am only going to
tell you about the
22 one
that is right. That way you will
remember that
23 I
got it right the first time. Wrong, but
24
nonetheless, we speculated that it was the protease
25
ClpP that was responsible.
241
1
[Slide.]
2
It is a complex protease that involves
3
ATP. You do the hypothesis-driven
experiments, you
4
knock it out, and as you can see by the figure on
5 the
righthand side, you eliminate the
6
oxygen-dependent proteolytic cleavage, not all the
7
cleavage, but that hypothesis, which was one of
8
about a dozen that we explored, in fact, worked.
9
[Slide.]
10
Are there other ways to think about these
11
kind of problems? Do we have
analytical techniques
12
that allow us to probe much more broadly the global
13
cell response?
14
[Slide.]
15
Of course, the answer is yes, and the
16
technology of using DNA microarrays to do
17
transcriptional profiling is one kind of tool that
18 can
be used in identifying where the problem is,
19
and, after all, isn't that what PAT is about.
20
It is about getting at the underlying
21
science to understand what the issue is, and then
22
focus on the right issue, not necessarily measuring
23
everything that you possibly can measure.
24 E. coli is very convenient. It only has
25
about 4,000 genes, but fortunately, those genes are
242
1 set
up in pathways, and rather than think about
2
4,000, I don't like big numbers, I would rather
3
think about, well, there are about 170 pathways.
4
[Slide.]
5
So, if we look at the response in terms of
6
pathways, we can begin to say, well, are there
7
pathways that are up or down-regulated, and,
8
indeed, this is an example in the case of the
9
experiments I showed you a moment ago.
10
What you see on the lefthand side is the
11
regulon associated with the peroxide response for
12 E.
coli to high oxygen--excuse me--on the lefthand
13
side to the superoxide response, the righthand side
14 is
the peroxide response.
15
What you can see by the elevated levels of
16 the
genes in the superoxide response, that E. coli
17
reacts with the operon, superoxide dismutase and
18
some other enzymes, and the peroxide response is
19
transient, if anything at all.
20
This tells us where the problem is.
The
21
problem is associated with the small amount of
22
superoxide radical that is being made.
23
[Slide.]
24
When we look at clusters of genes, one can
25 see
that the green ones are up-regulated in the
243
1
presence of oxygen, and when they are red, they are
2
down-regulated, which happens in the case of
3
nitrogen, and you see green dots with superoxide,
4 but
one of the other strange things is that you see
5
proteins that have iron/sulfur in them
6
up-regulated.
7
Why would any self-respecting E. coli
8
up-regulate genes associated with iron/sulfur
9
proteins when you are making a recombinant protein?
10
This didn't make sense, and, in fact, it has
11
nothing to do with the production of alpha-1
12
antitrypsin, but rather has to do with the fact
13
that a small amount of superoxide, that free
14
radical, knocks out the iron/sulfur clusters.
15
There are about 100 proteins in E. coli
16
that have them. Those proteins
are not functional,
17 so
how does the cell respond? It
up-regulates
18
pathways in order to compensate.
19
[Slide.]
20
So, within these global techniques, you
21 can
begin to understand where the problem is and
22
think about the strategies to better design the
23
process, and basically, it is about taking the next
24
step.
25
Where is the appropriate next step?
I
244
1
will leave it to your imagination whose feet they
2
are.
3
[Slide.]
4
A little quick self-assessment.
When we
5
introduce a process to make a biotherapeutic
6
product, do we know the optimum conditions for
7
quality and quantity of the product today?
8
No, and as a consequence, once the process
9 is
in place, we see very substantial process and
10
product improvement during the course of operation,
11 and
that is good because it means that we have
12
recognized that there is going to be variance and
13
that we have recognized that we can manage that
14
variance, we can learn from it, and collectively
15
benefit. That is the reality.
16
So, there are lessons learned there.
The
17
variance that is going to occur is not something to
18 be
avoided, it is something to embrace and learn
19 how
to manage, and it is getting the right balance
20 of
managing that risk.
21
So, during routine manufacturing, do we
22
improve the product in the process?
Absolutely.
23
[Slide.]
24
What is the way forward? Well, is
there a
25
better way than incremental adjustments to optimize
245
1 and
scale a process? Sure, and I think the
idea of
2
taking these complex processes and learning how to
3
operate large numbers to capture design of
4
experiments and to capture what happens in that
5
space, and learn how to assess the
6
interdependencies of the parameters is a very
7
exciting opportunity.
8
The technologies that allow us to do it,
9
both from a process side, from an analytical side,
10
from a data analysis side are really important to
11
bring together, and we are not there, but we can be
12
there.
13
We need to live with variance and take an
14
adequate opportunity to learn from that variance.
15
Listen to the data, don't ignore it, listen to it.
16
In doing that, we can again grasp much
17
more experimental space both in variables, as well
18 as
time. So, this issue of embracing that
19
variance, learning what it is about, learning where
20 the
problem is, and then using that to come back
21 and
develop a robust process, this is the kind of
22
mind-set that PAT is about, and biological
23
processes are very much in need of being thought
24
about and treated and respected in this way.
25
[Slide.]
246
1
In closing, the last slide is to look at
2
what are some of the process evaluation tools.
3
This is not all-inclusive, but it is meant to
4
reinforce just a couple of points that I have made.
5
One is the leverage analytical technology
6 on
process and products, what does this really
7
mean? This is PAT, and this is
leading us to a
8
process understanding and a process evaluation.
9
That is very much what it means, that we
10
need to be able to look at the process globally,
11 and
not just locally. It fits exactly in
with the
12
guidance that has been laid out for PAT.
13
We need to explore the biological space
14 and
the parameter variance. We need to
understand
15 how
this variance propagates through a process.
16 It is very interesting, if you take
17
process simulation tools, and we can do a very nice
18
process simulation on any of these processes, and
19
then you do things like Monte Carlo simulation
20
where you have variance in the process, you can
21
begin to understand how that variability at
22
multiple steps is going to propagate through very
23
complex processes.
24
As a consequence, when you do that, you
25
then are not surprised by how a little bit of
247
1
variance here, a little bit of variance there,
2
propagates to give you what the end result is going
3 to
look like. So, with simulation and these
tools,
4 you
can avoid some surprises.
5
We need to better interrogate the cell at
6 the
molecular scale, and then be able to do the
7
multi-scale analysis to scale up.
So, part of what
8 I
think PAT is about, is multi-scale analysis,
9
driving down to understand the science, so we can
10
understand where the problem is, and then driving
11
back up with appropriate solutions to eliminate the
12
right problem, in the right way, at the right time.
13
A lot of this about understanding these
14
interdependencies in what is a very large
15
experimental space.
16
Lastly, understanding this connection
17
between the molecular processes, process
18
performance, and product quality.
We are doing I
19
think an exciting job with drug substances in this
20
regard, and we are perfectly capable of carrying
21
that over, with work, to biological products, as
22
well.
23
I will stop there and I hope that I have
24 generated more questions than providing
answers,
25
because that is what I started out to do.
248
1
DR. KIBBE: Thank you. If there anybody
2 who
has any quick questions you want to take care
3 of
now before I go on, any point of understanding?
4
[No response.]
5
DR. KIBBE: In that case, Dr.
Koch.
6
DR. KOCH: I have had the benefit
today of
7 a
number of speakers who were leading up to the
8
type of things that I wanted to say.
I left out
9
some things, and those of you who have paged
10
through the slides probably can't believe that.
11
There is a lot of slides there, it is going to be a
12
little bit like a fire hose here for a while. I am
13
going to try to stick to things that are more of a
14
miniature nature or micro-analytical rather than
15
hitting the broad base of all analytical.
16
Let me move into it and I think I will tie
17 in
with some of the previous speakers.
18
[Slide.]
19
PAT. We have heard a number of
20
definitions of it, but again it is looking at all
21
aspects from the chemistry tools through the
22
control strategies and into the data handling
23
aspects. The goal again, process
understanding.
24
[Slide.]
25
The origin of PAT goes back, oh, 50 years
249
1 at
least, and we have got a few examples that go
2
back to the mid-forties with some of the German
3
chemical companies applying it, so it is not as if
4 the
approach is new.
5
We can go into all of the reasons why it
6 is
relatively new in the pharma industry, but that
7 is
mostly psychological. It started within
the
8
analytical chemistry labs where tools used for
9
specifications, et cetera, as coming from the areas
10
listed here, were then made portable for running in
11 the
process or close to where the process was, and
12
adopting the term "real time analysis."
13
[Slide.]
14
That real-time data resulted in a number
15 of
things, in fact, almost every time one went into
16 a
process, and this is borrowing from the
17
petrochemical experience, almost every time a
18
sample was taken to a chemical analysis lab, we
19
found out that the results were different if we did
20 it
in real time, taking and watching things that
21 you
could see fleeting intermediates or a number of
22
things that were indicating both safety and
23
environmental problems.
24
It also was a very good scoping tool for
25
understanding what type of issues and what places
250
1 in
a process could be monitored, process
2
understanding results from doing this.
3
[Slide.]
4
What is appropriate for PAT? It
really
5
comes down to a very broad statement, and that
6
anything that gives you data that you are presently
7 not
measuring, certainly want to look at cheaper
8 and
more reliable, and then we are entering into
9
something here where we are going to get more data
10
than we ever wanted, but we are going to want
11
additional data points in order to build better
12
models from which to control from.
13
This is probably going to be the crack in
14 the
wall for Bayesian type approaches where you
15
have to make assumptions because you finally get
16 too
much data that you can't possibly study all of
17 it.
18
It is also going to allow us to depart
19 from
traditional analytical science technologies,
20
that list that showed up before as coming out of
21 the
analytical laboratories, have to move away from
22
that.
23
[Slide.]
24
We are going to have to look at fully
25
integrated analyzer systems.
Historically,
251
1
analysis is detection. The thing
that people have
2
avoided forever is the problem with sampling, I
3
think taking inadequate representative sample to be
4
analyzed.
5
Then, the other thing that often was
6
slipped over because of expense and capability had
7 to
do with collecting the data and making sense out
8 of
it, and eventual information and knowledge.
9
That has to be all integrated into a system.
10
The next point has to do with inferential
11
analysis, and we have heard that referred to a
12
couple of times, and that is where you can project
13 to
the desired product properties by doing some
14
measurement during the process, and it doesn't have
15 to
be the property itself, but you have enough data
16
that you can extrapolate to that point.
17
[Slide.]
18
Then, you have to revisit some of these
19
underutilized, but not revolutionary techniques.
20 The
few that I mention here are technologies that
21
were discovered in the early 1900s, but not used
22
forever, largely because of instability of optics
23 or
computer possibilities back when it was first
24
looked at.
25
[Slide.]
252
1
I want to mention a couple of these.
One
2 is
in the optical low coherence reflectometry, that
3
when you do that type of measurement, the result
4 you
get depends on things like on the column on the
5
left, the thickness, the particle size,
6
concentration, shape, and some of these other
7
morphological things all affect the measurement.
8
As a result, if you can interpret the
9
signal that you get from the measurement, you can
10
then use it to monitor a number of things. There
11 are
examples there, that are largely from a
12
chemical and materials point of view, but
13
eventually, you get down to being able to monitor
14
tablet coating.
15
The technique started in measuring coating
16 of
airplane wings, and we found that that could be
17
extrapolated quickly to other measurements that is
18
being used now for tablet coating, as I mentioned,
19 and
we are finding that there is variations during
20 a
fermentation or a biological process that can be
21
monitored, and it is a technique that operates at
22
high concentration, in slurries of 70 to 80 percent
23 as
a technique for particle size versus the
24
historical need for dilution.
25
[Slide.]
253
1
A couple of examples. You can
look at a
2
multi-layer film. Here is an
example of a drug
3
delivery patch. I think you can
see some of the
4
peaks there on the bottom.
5
Very interestingly, what happens in this
6
process, it looks like a chromatogram with various
7
peaks, however, it is the bounce back of the
8
photons at each layer, and you measure the time
9
that it takes to come back and project into
10
distance.
11
Each one of those peaks is a layer.
It is
12 a
layer from the barrier layer on the outside and
13 the
back, and then the intermediate layer is
14
between active ingredients, so it becomes a way to
15
measure how much active ingredient one has placed,
16 so
the baseline is basically the thickness of the
17
active ingredient.
18
The scattered material example is one
19
where you have a total reflection of the photon and
20 the
path in which it travels indicates the
21
complexity of the mixture, and you can extrapolate
22
then into things like particle size, shape, and
23
waveguide formation, et cetera.
24
[Slide.]
25
An example of being able to look at
254
1
consistency, there is one curve here that shows at
2 one
concentration, you can see quite a range of
3
small particles from basically 20 to 90 nanometers,
4 or you can take one size, in the lower
example, of
5 308
nanometers, and get a concentration difference.
6 So,
it has proven to be quite valuable in that
7
regard.
8
[Slide.]
9
Moving on to Raman, certainly, everyone
10 has
heard the terminology, but as you look at some
11 of
the potential advantages now that the stability
12 of
the lasers have improved in some of the data
13
handling, and as databases grow, you can look at
14
non-invasive or non-destructive technology.
15
You can work in aqueous systems.
You can
16 do
multiplex of your instrument using fiber optics
17
that can go hundreds of meters, and you can also
18
then look at chemical structure and fingerprinting
19 of
both inorganic and organic materials.
20
[Slide.]
21
Then, with effective probes, in fact, this
22
particular probe that is demonstrated here, was the
23 one
that we used in the practicum and moved between
24 the
various centers to study some milling and
25
mixing operations, but we have done a number of
255
1
things in composition, as well as, at the bottom
2
right, putting it in a protein mixture in terms of
3
determining aspects of that material.
4
[Slide.]
5
The fringing electric field or
6
dielectrometry sensor is pretty simple, one that
7 was
developed for detecting mines, and it has to do
8
with the ability to set your electrical fields with
9 the
various sensors in setting the distance and the
10
intensity, and you can get a disturbance of that
11
electrical field based on the properties of the
12
sample.
13
You can measure things like density,
14
distance from the sensor, texture, and moisture,
15 and
moisture not only in concentration, but
16
distribution, so you will start to look at filter
17
cakes or other aspect of various processes. You
18
have another relatively unused method that can be
19
applied.
20
[Slide.]
21
To date, a number of things happening in
22 the
paper pulp industry, pharmaceutical products,
23 and
we have got a few companies, pharma-based, that
24 are
using it for mixing consistency, a lot of food
25
applications including some of the baking companies
256
1 to
monitor the moisture distribution in cookies and
2
cakes and things, and that turns out to be pretty
3
important for them, composites, plastics, et
4
cetera.
5
[Slide.]
6
Going on to surface plasmon resonance, a
7
number of things, primarily in miniaturization and
8
sensitivity have occurred here, and plugging some
9
disciplines together from electrical engineering
10 and
genetic, have come up with some real-time
11
biosensors that are operating at a very fast mode.
12
[Slide.]
13
Work sponsored by the Department of
14
Defense, again that tie in with some of the things
15 we
heard earlier on homeland security.
16
You can start to look at high throughput
17 screening,
automated protein purification, and
18
number of toxins, food-related activities, and we
19 are
actually moving quite rapidly into response in
20 the
food industry for safety, security, nutrition
21 in
the food and related water chains.
22
[Slide.]
23
One example, this has been demonstrated in
24 a
protein purification system, would be a way in
25
which after the broth is separated and some
257
1
chromatography applied using biosensors, one can
2
determine when to change columns or monitor the
3
process, which brings us to biosensors and the need
4 in
the bioprocess in general.
5
[Slide.]
6
I have been in discussions
with Harry Lam
7 of
Genentech, to get a feel for what type of things
8 the
industry is looking at, and certainly to
9
maintain a consistent product performance or
10
process performance with the development cycle from
11
early stage through manufacturing.
12
[Slide.]
13
Measurement is needed in order to look at
14 the
underlying functional relationships that occur
15 in
the process, as well as some of these
16 interactions
of the organisms with their
17
environments.
18
[Slide.]
19
We need to improve the capabilities for
20
process control, and the type of measurements are
21
going to be broad based, biological, chemical,
22
physical.
23
[Slide.]
24
Much of this has been touched on today -
25
biological with this whole range of things that are
258
1 of
a cellular nature.
2
[Slide.]
3
Chemical, we have got a number of things
4 in
the media that need to be addressed, that have
5 to
do with the nutrients and the additives, et
6
cetera.
7
[Slide.]
8
It continues on when you start to
9
characterize the product, the by-products, the
10
environment, as well as the off-gas.
11
[Slide.]
12
Physical. We have heard much of
this in
13
terms of the type of things that need to be looked
14 at.
15
[Slide.]
16
What can we look at today? Much
of this
17 was
mentioned here in the last couple of
18
presentations, of things that are being used to
19
monitor, but that leaves a number of the issues on
20 the
table yet to be addressed and solved.
21
[Slide.]
22
Also, the industry is looking at the
23
various requirements that are going to be
24
necessary, and it is a lot more than just having a
25
measurement tool, but to get into the things that
259
1
have to do with sterilization, interference, and
2 the
fouling, low maintenance, and the small size.
3
We have heard several assumptions today
4
that if it's smaller, it could be better.
5
[Slide.]
6
That gets us into what has been driving
7 the
improvements in measurement over the last, say,
8 20
years, and it has been the advances in
9
miniaturization. Much of this has
been driven by
10
technologies in the computing industry and the
11
ability to make things smaller and use microfluidic
12
technologies, et cetera.
13
Certainly, new materials, the optic
14
advances, and computing have helped, but
15
miniaturization is really a big one.
16
[Slide.]
17
It has been focus of the center where I am
18
located in Washington. It has
been a
19
multi-industry, and I have implied that a few
20
times. A number of industry come
together and
21
discuss advance in real-time measurement, and we
22 are
now beginning to apply those things to the
23
food, pharma, biotech industry.
24
[Slide.]
25
Multidisciplinary. There are many
260
1
examples where bringing different disciplines
2
together results in some very interesting sparks
3
coming from that smoke, presently supporting 20
4
different research projects at 5 universities, and
5
involved with some international collaboration.
6
[Slide.]
7
The initiatives, and we will see the
8
importance of this growing, is sampling and
9
sensors. That is one that we try
to act as a forum
10
across industry. Trying to also
compile analytical
11 and
chemometric methods, what to use in terms of
12
interpreting the data.
13
A couple of things that are used just
14
inside for the members, are to look at
15
micro-instrumentation for the high throughput
16
experimentation, the CombiChem, and some of the
17
process optimization tools, and then a fermentation
18 platform,
and I will mention some of the things
19
there.
20
[Slide.]
21
When we look at this response to high
22
throughput experimentation, we get into the
23
micro-instrumentation world, but also the
24
micro-reactor world. I have to
agree with that
25
Charles mentioned, the petrochemical industry is
261
1
finding huge benefits in scaling down before you
2
scale up, and going down to molecular interactions
3 in
a number of data-gathering aspects at the small
4
scale to understand how to then move on from that
5 to
macro scales.
6
[Slide.]
7
We also have a number of techniques that
8 are
being miniaturized largely due to advances.
As
9 I
mentioned before, most of the analytical
10
technologies, we only have a few that have not been
11
miniaturized yet or taken on-line, and some of
12
those are microscopy-based, but we actually have
13
some breakthroughs now in bifringes and other
14
things that could help in this respect.
15
[Slide.]
16
I will give you a couple of examples.
In
17
micro-LC, we have got a small 100-micron flow
18 channel where you mix a sample at a mobile
phase
19 and
then detect the deflection in your laser beam
20
with a position-sensitive detector.
21
[Slide.]
22
We have since found, after starting into
23
this project, that low molecular weight material
24
diffuses much faster than the higher molecular
25
weight material.
262
1
[Slide.]
2
So, why not put two sensors in-line and
3
then begin to calculate the difference between
4
those distances in terms of a particular molecule.
5
[Slide.]
6
What has resulted is an in-line molecular
7
mass sensor where we are able, in this case, to
8
look at polyethylene glycols from a very low
9
molecular weight. Actually, it has now been taken
10 to
over 100,000 molecular weight in terms of a
11
standard curve.
12
[Slide.]
13
This has resulted in other things now, in
14
some biological testing where we can see peptide
15
synthesis, we can look at polysaccharide synthesis
16 and
be able to see differences as chains are
17
building, and also be able to see differences in
18
diffusion in following trends in that way.
19
[Slide.]
20
Developments at Sandia, again, homeland
21
security basis, have resulted in a micro chem lab.
22
This is a very interesting thing, obviously, the
23
size of a dime is quite impressive, but when you
24
look at the SAW ray detector you have go a 1-meter
25
column, and your sample pre-absorption.
263
1
[Slide.]
2
You put all that into a hand-held unit,
3
this is a now a hand-held GC, but the end there
4
indicates it is also an LC, so that has all been
5
incorporated into taking today's lab technology
6
down to a very small size.
7
[Slide.]
8
Some work that we have been involved with
9
recently is when you go to use of nanoparticles in
10
your column, you can increase the speed.
We are
11 now
talking of these compounds being separated in
12 two
seconds.
13
Normally, you are looking at 40-minute
14
type turnarounds on a lot of these GC analysis
15
things that have been improving, and I can't really
16
talk about it, but we now have a similar separation
17 in
500 milliseconds, that things are really flying
18 in
that way, so it has become a real-time
19
analytical technology.
20
[Slide.]
21
A small mass spec has been developed.
22
There is three or four examples of taking mass spec
23 down
to these small sizes.
24
[Slide.]
25
We are also involved with development at
264
1
UC/Davis with the micro labs, the electrical and
2
computing and the food science areas, to develop a
3
NMR.
4
[Slide.]
5
This is an NMR now that early signal is
6
shown on the bottom left, which showed water, a lot
7 of
excitement by the food group because they could
8 monitor a number of things in real time. It
has
9
since been refined to the bottom right there, and
10 it
has been taken from a protein signal, we have
11 now
seen carbon and phosphorus, so we are talking
12
about a hand-held NMR that is going to be
13
multinuclear and have a cost of probably under
14
$20,000.
15
[Slide.]
16
So, all these advances in sensors and
17
controls again highlight the need, how do you get
18 the
right sample to these technologies.
19
[Slide.]
20
The chemical industry has come to us, and
21 we
have been a forum for discussions on how to
22
create new sampling and standardized technologies
23 in
that arena.
24
[Slide.]
25
The typical sampling in a petrochemical
265
1
plant is a large, often covering a wall, quarter of
2 a
million dollars worth of instrumentation just to
3
interface the process with the analyzer.
4
[Slide.]
5
That has now shrunk down to an inch and a
6
half by an inch and a half modules, a standard set
7 by
the ISA, and this platform now houses the valves
8 and
filters and regulators to interface again the
9
process with the analyzer.
10
[Slide.]
11
What has been evolving here, this concept
12
started in late 2000, and it has now generated to
13
point where we are beginning to think of how we
14
could make this Smart and how to utilize advances
15 in
micro-analytical.
16
[Slide.]
17
So, the base here has been defined.
We
18 now
have a standard sampling interface that can be
19
heated or cooled, or whatever, and the flow
20
patterns all defined, and in the next couple of
21
months, we are standardizing a connectivity.
22
This is getting into some control
23
engineering terminology of how do you move the
24 signals
from that platform to distributor control
25
systems and other fields of how do you use that.
266
1
[Slide.]
2
Then, what has happened is you can now
3
drop your pressure regulators, your valves, and
4
your filter onto that platform and be able to
5
monitor what they are doing.
6
A very interesting story happened at again
7
interfacing the process with the analyzer. The
8
first year of use of these devices caused the
9
engineers to say why does the analyzer have to be a
10
refrigerator size, when the sampling system has
11
come off the wall to this fairly small
12
compartmentalized unit.
13 So, this platform has now become the
base
14 for
micro-analytical, so it has become a standard
15
platform for the development of micro devices.
16
Three or four years ago, if somebody had
17
come in with a small GC and say wow, isn't this
18
neat, and we would say that is really nice, but how
19 do
we use it, how do we go to this big,
20
wall-mounted sampling system and put this little GC
21 at
the end of it.
22
That has changed, people are now putting
23 on
a fair amount of suction for the development of
24
these devices.
25
[Slide.]
267
1
So, we predict, and it is beginning to
2
happen, that the NeSSI platform will become the
3
base for a micro-analytical lab.
Already we have
4
oxygen and pH and moisture, mass flow controllers,
5
little mass specs, all of the techniques that are
6
listed there have the plan to be mounted on this
7
particular platform.
8
[Slide.]
9
And then we have been devising different
10
interfaces. Our Raman sensor now
will fit on the
11
NeSSI platform.
12
[Slide.]
13
The surface plasmon resonance, this is the
14 one
that does the very fast biological detection,
15 is
now down to the size where the flow channels
16
will interface with the surface and provide almost
17
real-time biological detection in the NeSSI
18
platform.
19
[Slide.]
20
And we have taken something that basically
21
used to be flow injection analysis, it migrated to
22 be
called sequential injection analysis, now it is
23
micro sequential injection analysis, where you can
24 put
wet chemistry on a multi-functional,
25
multi-position valve, so you can scale down wet
268
1
chemistry and titrations and things, and do things
2
like glucose, nitrogen, nutrients, and inorganic
3
detection, and this is now on the NeSSI platform.
4
[Slide.]
5
So, there is almost nothing right now that
6 we
don't have that couldn't possibly fit on here,
7 and
we see it, not only for the process control,
8 but
all kinds of optimization studies that could
9
interface with lab-based fermentation and with the
10
micro-reactor systems for the chemical world.
11
[Slide.]
12
The last thing I will
mention is our
13
Fermentation Initiative.
14
[Slide.]
15
We are trying to apply the known
16
techniques and compare them with things that are
17
evolving and have applications in other fields. We
18
want to provide training and understanding the
19
implication of some of these measurements.
20
[Slide.]
21
We have set up some platforms that are now
22
outfitted with this array of instrumentation, and
23 you
can see things like dielectric spectroscopy,
24 the
surface tension, light reflective spectroscopy,
25 et
cetera, that are not traditionally being used in
269
1 the
fermentations, but we are gathering data and
2
finding then ways to extrapolate to which
3
fermentation areas they will best influence, and
4
then looking at sampling.
5
Sampling and fermentation is a big
6
problem. Go back to some of the
things that we put
7
together that Genentech summarized.
8
[Slide.]
9
They are very concerned about how to
10
achieve these type of considerations, and then you
11 get
into some of the sterile requirements and how
12 do
you design your sampling system, so it will meet
13
these requirements.
14
[Slide.]
15
So, what we have is a plan to continue to
16
scope out activities from an analytical point of
17 view,
but to implement and evaluate this NeSSI
18
platform for not only sampling, but sensor and
19
process control interfaces, so we have a platform
20 now
that is being put in to work with sampling the
21
broth and another one with head space.
22
We are looking at chemometric tools to
23
model this batch variability and look at various
24
data fusion approaches. We need
to do this to
25
begin to develop these automated tools to evaluate
270
1
production data and implement chemometrics as much
2 as
possible for quantifying process performance and
3
applying these PCA approaches to performing
4
automated pattern recognition.
5 [Slide.]
6
To borrow a little bit from Helen's
7
earlier slide, it is going to be an exciting time.
8 We
have got a lot of things in front of us, but to
9
take the advances in PAT from the other industries,
10 through
the pharmaceutical on to the biological, I
11
think is going to be very rewarding.
12
DR. KIBBE: Thank you.
13
Are there any quick questions?
14
DR. SINGPURWALLA: I have a
comment.
15
You have this nice chart on
16
multidisciplinary, page 16, and you also had CPAC
17
initiatives. Just from a
parochial point of view,
18 I
noticed the absence of a statistician, yet, you
19 are
discussing the sampling, which is really a
20 statistical
issue.
21
DR. KOCH: You are right. In fact, that
22
list is not complete. What that
list is, is the
23
present principal investigators involved with our
24
programs. We have just finished a
project with the
25 chairman
of our Statistics Department where we were
271
1
funding things just for what you are saying.
2
So, really, that list is project
3
dependent. I could probably add
as many as six
4
other areas, like physics was on there, and a few
5
others in the past year, so we rotate projects in
6 and
out. Chemometrics is probably based in
7
statistics. They don't like to
admit it.
8
DR. KIBBE: Anything else? Ajaz doesn't
9
want to comment? Okay.
10
Tom, wrap us up.
11
DR. LAYLOFF: Much of what I
wanted to say
12 has
been said already, so I will speed through my
13
slides, and I have a few comments at the end that
14 are
not in the slides.
15
[Slide.]
16
First of all, PAT, with the subcommittee
17 to
this committee, advisory committee, we had a
18
series of charges which were given to us.
19
[Slide.]
20
We had meetings lasting through 2002,
21
three meetings. We covered
applications and
22
benefits, process and analytical validation,
23
chemometrics, process-product development, process
24 and
analytical validation, a proposed PAT training
25 and
certification program, which I think was one of
272
1 the
highlights of the activities.
2
Computer systems validation, 21 CFR 11,
3 and
Joe out there tackled that one, PAT case
4
studies, and rapid microbiological testing was
5
tacked on near the end there.
6
[Slide.]
7
We reported that back to this committee
8
back in October. There was a
definition of process
9
analytical technology. I am not
going to read that
10 to
you again, you have already seen it.
11
[Slide.]
12
Again, more statements on PAT
13
applications.
14
[Slide.]
15
Now, this was not included.
Historically,
16
there has never been anything to stop people from
17
using new technologies. As a
matter of fact, in
18 the
1978 preamble to the CGMPs, there is no
19
prohibition in the regulations against the
20
manufacturing of drug products using better, more
21
efficient, and innovative methods.
22
It is a big box, and it has been there
23
since 1978.
24
The USP also allows alternative methods
25 for
assessments.
273
1
[Slide.]
2
The committee proceeded by coming up with
3 a
general guidance through Raj Uppoor and the OPS
4
staff to generate a guidance, which defined a
5
regulatory position for the process and added some
6
incentives. Then, the FDA PAT
team, which came
7
through the training program you have heard about
8
earlier.
9
[Slide.]
10
Of course, the Agency's perspectives.
One
11 of
the things that I think is very interesting,
12
coming from many years of service in the Agency,
13 was
the Agency's use of existing knowledge,
14
experience, and guidances from other FDA
15 components, and NIST, ASTM, and ANSI.
16
The FDA tended over the period of my
17
tenure to be very introspective, if it wasn't NIH,
18 not
invented here in FDA, we had very little use
19 for
it. Going to ASTM and NIST was more of
an
20
engineering approach, and we tended to hang with
21 the
pharmacists in the USP.
22
The USP, of course, was established by
23
practitioners as a book of recipes to assure
24
quality, and we hung with that, with the
25
practitioners rather than with the engineering.
274
1
ASTM was established by engineers and
2
chemists to deal with defective rails in the
3
railroad, so they tended to be very
4
engineering-oriented, and it was quite interesting
5
that when FDA-CDRH went out looking for standards,
6
they went to the engineering standard type area
7
rather than practitioners of pharmaceuticals.
8
Now, the switch in OPS of looking
at ASTM
9
standards is very interesting because it moves
10
process analytical technologies into an arena where
11
there are engineers and chemists, scientists rather
12
than practitioners. It's a switch
in philosophy.
13
Also, the ANSI and ISO fit in that also.
14
They established a framework for manufacturers with
15
flexibility needed to develop new designs.
16
[Slide.]
17
Future issues. Validation data
and
18
retention. We have heard some
about retention of
19
data, and I don't think that has been addressed
20
well, but the process analytical technology is
21
going to deluge with information, and there is
22
going to have to be some way of defining what is
23
essential and should be retained, and what is not
24
essential.
25
The definition of in-process endpoint
275
1
detection, data acquisition and storage.
In a
2
process of PAT, you have to have some component in
3 the
process which is measurable and defines an
4
endpoint. You have to have
analytics, but you have
5 to
have something that you are looking for.
6 The documentation of the data acquired
and
7
electronic signature closures of decision points
8 are
going to be an issue, and the incoming material
9
stream consistency and robustness assessments are
10
going to be critical for supporting PAT also.
11
[Slide.]
12
Regulatory incentives, we have gone over
13
those already, not a requirement.
14
[Slide.]
15
How to move forward, try and do it by
16
evolution rather than revolution.
Don't bring it
17 all
up at once.
18
[Slide.]
19
And the guidance which came out in
20
September, just a few items from it.
21
[Slide.]
22
The guidance is intended to describe a
23
regulatory framework that will encourage the
24
voluntary development and implementation of
25
innovative pharmaceutical manufacturing and quality
276
1
assurance--manufacturing and quality assurance,
2
voluntary, innovative. Those are
key terms.
3
[Slide.]
4
The scientific risk-based framework
5
outline in the guidance should help manufacturers
6
develop and implement new and efficient tools for
7 use
during pharmaceutical development,
8
manufacturing, and quality assurance while
9
maintaining or improving the current level of
10
product quality assurance.
11
The framework we have developed has two
12
components: a set of scientific
principles and
13
tools supporting innovation, and a strategy for
14
regulatory implementation that will accommodate
15
innovation-keys.
16
[Slide.]
17
Among other things, the regulatory
18
implementation strategy includes creation of a PAT
19
team approach to the CMC review and CGMP
20
inspections and joint training and certification of
21 PAT
review and inspection staff.
22
The Agency is encouraging manufacturers to
23 use
the PAT framework described here to develop and
24
implement new pharmaceutical manufacturing and
25
quality assurance technologies.
277
1
[Slide.]
2
The guidance is written for a broad
3
industry audience in different organizational units
4 and
scientific disciplines.
5
To a large extent, the guidance discusses
6
principles with the goal of highlighting
7
technological opportunities and developing
8
regulatory processes that encourage innovation.
9
[Slide.]
10
Biologics and PAT. The umbrella
guidance
11
covers biological production within the scope.
12
Presentations before our committee included
13
individuals that were using, or companies that were
14
using, process analytical technology to monitor
15
fermentation and purification of biological
16
materials, so it fits if you can define those kinds
17 of
controls.
18
However, the process differences that
19
occur in biologics may require or likely will
20
require additional skills and an expansion of the
21
training and certification program.
22 So, that PAT concept, the training of
23
reviewers and inspectors will probably need to be
24
expanded with training in biologics type PAT
25
applications. The standard
chemical stuff is not
278
1
going to work, you are going to have to expand it
2
beyond that.
3
But I think the concept that came out of
4 our
committee of having a training program and a
5
certification of competencies is very useful for
6
building teams to get around some of the silos that
7 we
have in FDA, that have been there, because those
8
silos contribute to poor science and poor
9
regulation, some of them, and that needs to be
10
straightened out, and this is a good attempt at
11
beginning to do that.
12
[Slide.]
13
Acknowledgments. Ajaz has done a
great
14
job, I am a great fan of his efforts in taking this
15 and
driving it forward because he has really done a
16
great job of pulling it.
17
Raj for doing the guidance. My
former
18
colleagues at the DPA, DPQR, colleagues that
19
presented at the PAT Committee, and those reports
20 are
at that web site.
21
It has been a lot of fun for me to work on
22 the
PAT Subcommittee. As Ajaz said, it was a
23
project that we started about 11 years ago, and to
24 see
it come to fruition now has really been great.
25
I think the industry has got to do more,
279
1
they are going to do more as they reduce their
2
inventories and move to just-in-time manufacture,
3
which will reduce cost, and, very importantly, help
4 bring the vision of health to all closer.
5
That's it. Any questions?
6
DR. KIBBE: Any questions for
Tom? Now,
7 we
understood that you had answers to all of these
8
questions from Keith Webber.
9
DR. LAYLOFF: I do, I do.
10
DR. KIBBE: If you could just tell
us what
11 the
answers are, we could all go to--Happy Hour,
12
right.
13
No questions? Perhaps, Keith, you
would
14
like to lead us through these questions and try to
15 get
at least some of our collective wisdom on some
16 of
them.
17
Committee Discussion and Recommendations
18
DR. KIBBE: What technologies are
19
available now to evaluate the characteristics of
20
protein products in real time during manufacturing?
21 Who
has an answer?
22
Dr. Koch.
23
DR. KOCH: I don't know if I could
answer
24
that directly, but there are a number of monitoring
25
methods that are being used by those manufacturing
280
1
protein products today, and it probably might be
2
best to pool some kind of a compilation from those
3 who
are presently in that product arena.
4
Often from the place where I am sitting, I
5
have a difficult time judging what measurement
6
techniques that we develop, and we have been
7
involved with technologies that have made it into
8
commercialization, but when we ask one of our
9
members is it working or how well is it working,
10 and
their processes, we can normally tell just by
11 the
smile or lack of it.
12
So, we are in a situation where we are
13
developing tools for a toolbox, and we are never
14
quite sure how well they are being applied.
15
I don't know, Gerry, you are probably in a
16
position, or Rick, I saw earlier.
17
MR. MIGLIACCIO: My background is
on the
18
small molecule side, so I wouldn't want to leap
19
into this.
20
DR. LAYLOFF: I have a question on
what
21
does it mean evaluate the characteristics, because
22 the
sequence is pretty well clean on proteins.
Are
23 you
talking about secondary, tertiary, quaternary
24
structures?
25
DR. WEBBER: I am talking about
more the
281
1
overall structure, say, tertiary, and in those
2
cases where it is applicable, quaternary
3
structures, but also the post-translational
4
modifications structure of the product, which is
5 one
of the I think most variable in terms of what
6 we
see to change during, say, fermentation, or it
7 can
be selected out during purification.
8
You can get various species of product get
9
selected or rejected during purification, so that
10 is
really what I was looking at there.
11
DR. LAYLOFF: So, it is not a
process
12
closure, it is actually an assessment of the
13
product coming out of the process.
14
DR. WEBBER: In this particular
question,
15
yes, it is the product which is not necessarily
16
coming out, but the product during manufacturing.
17 One
of the areas that I have seen reported is the
18
ability to use--and I discussed it a little
19
bit--immunological techniques or lectins to look at
20
structures.
21
For example, carbohydrates on products, I
22
haven't seen that in practice yet as a PAT, but
23
that is something that may be coming down the line.
24
DR. KIBBE: Dr. Cooney.
25
DR. COONEY: First of all, this is
a very
282
1
important question because this is a question that
2
relates what you make to its therapeutic safety and
3
efficacy, on the one hand, so part of addressing
4
this question is to understand the relationship
5
between particularly post-translational
6
modification, glycosylation, acetylation,
7
phosphorylation, and so on, and its therapeutic
8
efficacy.
9
It is also important because a lot of
10
those properties are known to vary with the
11
process, so whatever you use at this point for an
12
assay or an analytical technology, links you back
13 to
the process, on one hand, presumably you will
14
understand that linkage, and links you forward to
15 the
patient, on the other hand.
16
I think the advances in mass spec that
17
have evolved with proteins and, in particular, with
18
proteins that are modified, is quite substantial.
19
Quite recently, I saw ultra-high pressure
20 chromatography,
which takes advantage of a number
21 of
innovations in chromatography by being able to
22 go
to very small particles with a very high amount
23 of
surface area at very high pressures, and by
24
doing that, you can very quickly get a very high
25
resolution of complex mixtures.
283
1
So, a combination of size, shape methods
2
plus mass spectrometry and being able to work with
3
large molecules and very small amounts of material,
4
that seems to be where things are going, and
5
provides a very powerful armamentarium of PATs.
6
DR. KIBBE: Tom has another
comment?
7
DR. LAYLOFF: I was wondering, on
8
chromatographic procedure, whether or not you could
9
have subsequent post-translational modification of
10 the
proteins themselves. Denaturation would be one,
11 but
reactions with the supports themselves at
12
15,000 psi are reactive with solvents catalyzed on
13 the
supports.
14
I don't know how you validate the
15
separation tools at 15,000 psi.
16
DR. COONEY: You raise a very good
point.
17 The
work on ultra-high pressure chromatography is
18
very new and solvents and dissolved gases and
19
solvents are very reactive at those pressures.
20
That needs to be sorted out. It's
the right
21
question, and I think one can design the
22
experiments to get the answer.
23 DR. KIBBE: I feel like we are doing 1 and
24 2 a
little bit. What tools would allow us to
25
understand the manufacturing process better? The
284
1
tools that will allow us to understand the product
2
will also allow us to go back and look at the
3
process.
4
Is there anything specifically that
5
anybody would like to add on that?
6
DR. LAYLOFF: I was going to say
one of
7 the things that comes up, of course, in near
8
infrared, in applications, is that you don't have
9 to
separate anything, you just look at it, and you
10
define your endpoints on a polyvariate system.
11
It may be also possible to do something
12
like looking at a mass spec fingerprint without
13
separating anything, just look at the mass spec,
14
just hammer it and see what it looks like during
15 the
course of a process, just hammer it at
16
intervals and just see what it looks like until you
17
define an endpoint by another source, and then use
18
that as an endpoint indicator.
19
DR. KIBBE: Anything else you
need?
20
DR. WEBBER: Just one follow-up
question
21 with
regard to No. 1 from the presentation that Dr.
22
Koch did.
23
You had shown LC and NMR technologies that
24
were miniaturized, and we like to think smaller is
25
better and more PAT-like. Would
those
285
1
technologies, as they are now, be amenable to
2
biotech products or do smaller molecules, would
3
they be useful for looking at, say, fermentation
4
components, and things like that?
5 DR. KOCH: We plan to have all those
6
techniques tied in with the fermentation project,
7 so
there are early reasons to believe that we will
8 be
getting data from them.
9
I think one of the important things to
10
point out, when we even talk about DNIR or Raman or
11
some of the others, I think we will find with time
12 an
array of technologies with a multivariate
13
evaluation of the data is going to prove in the end
14 to
be quite valuable, so that you can look at, and
15
see, the variations that are coming from batch to
16
batch or system to system.
17
DR. WEBBER: Thank you.
18
We had completed Item No. 2 or not, you
19
sort of led into that, but are there any other
20
comments with regard to what tools would be
21
available to allow us to better understand biotech
22
processes?
23
DR. KOCH: Maybe just a comment on
that
24
one. It seems like at the top of
most
25
manufacturers' list is bioviability, and that takes
286
1 on
all kinds of definitions based on what product
2 one
is working with.
3
The more tools that are developed to
4
determine the health of the organism, the maturity
5 of
the system, or measurement or metrics to
6
determine when is the best time to harvest, there
7 is
a number of things I believe are going to be
8
advancing there, both direct and indirect methods.
9
DR. LAYLOFF: Then, there is also
going to
10
probably be indirect methods, like on flowing
11
stream systems, where you actually take the
12
fermentation broth, react it on to other species,
13
which could serve as a surrogate to where the
14
process is located.
15
DR. COONEY: One of the things
that will
16
surely happen in the diagnostics field is improve
17
proteomic techniques. The genomic
techniques are
18 not
so bad, the proteomics are still early stage,
19 but
as we develop better proteomic techniques, as
20 you
develop better immuno-based panels that are
21
important in diagnosis of disease, there is going
22 to
be a spillover benefit to the application of
23
these to the processes themselves.
24
So, this not a static, obviously a static
25
situation, and I expect that the main driver for
287
1
some of the new analytical techniques will not be
2
process understanding, but rather will be
3
understanding the biology, and that it is up to us
4 to
take those same techniques and those same
5
methodologies and begin to apply them to the
6 processes.
7
The other piece of this, to emphasize a
8
point I made earlier, by being able to do a lot of
9
measurements on a small scale, one can take
10
advantage of experimental design and look at your
11
experimental space, so these techniques that allow
12 you
to do that are important.
13
Another area, I mentioned doing
14
large-scale fermentation type of experimental
15
programs, but you need to do this for downstream,
16 as
well, and there is a fair amount of work, there
17 is
a modest, well, there is a little bit of work
18
being done to miniaturize the downstream processes
19
that hopefully should have the benefit of also
20
being able to do design of experiments on a larger
21
amount of downstream space at the same time, so
22
there is yet another area of development
23
particularly in the microfluidic space.
24
DR. KIBBE: Just a quick follow-up
25
question of our experts over here.
One of the
288
1
things I have noticed whenever we discuss PAT, is
2 we
deal with a tremendous amount of data influx, we
3 get
lots of data, and then we have to sort out the
4 data
that is really valuable to us.
5
Is there a role to play for the
6
ever-increasing power of the computational machine
7
that sits next to the instrument?
8
DR. COONEY: Absolutely yes, not
only in
9
working your way through large data sets, but also
10
learning how to do simulation both at the molecular
11
scale and upwards.
12
There is very interesting work being done
13
with modeling of small molecule-protein
14
interactions that is useful from a design point of
15
view, but it is also useful to explain some of the
16
phenomena that you see in a process.
17
So, having large computational capacity is
18
very important both from the passive data mining,
19 as
well as the proactive process simulation role.
20
DR. KOCH: I have to more than
second
21
that. The number of sensors being
developed, and
22
that is just begging for a number sensor mining and
23
then into the data mining, and then on into how do
24 you
handle the monstrous amounts of data.
25
DR. KIBBE: Just a personal
opinion about
289
1
accepting monstrous amounts of data is I sincerely
2
hope that the companies analyze it, pick out what
3 is
important, and the FDA accepts only those things
4
that are worth looking at, and doesn't demand every
5 ton
that comes through the door.
6
DR. SINGPURWALLA: Well, I have to
7
disagree. I am suspicious of data
mining because
8 you
are looking at patterns. You may look at
9
patterns that are purely imaginary.
There is a
10
classic example of consumption of alcohol and
11
professor's salaries. You know,
there are dubious
12
correlations that can come about.
13
Now, having said that, I think when you
14 are
exploring any data--and I think you asked a
15
very good question, and I am not sure if the
16
question has been addressed--with a lot of data, we
17 are
collecting a lot of data, by itself, may not
18
contain the knowledge of the information that you
19 are
really looking for.
20
You may collect a lot of data which
21
provides information which is not really relevant
22 to
what it is that you are interested in.
So, the
23
whole idea is when you are doing a data analysis
24
rather than data mining, what you have to do is
25
have some kind of a hypothesis in mind, have some
290
1
kind of a model in mind, and the model is never
2
suggested by the data, the model is always
3
suggested by the science that you are looking at
4 and
let the data then give you the unknowns of the
5
particular model or help you change your model or
6
help you update your model.
7
So, I think your question is very nice and
8
very important, and I think it goes back to why
9
collect the data. You should have a purpose for
10
collecting the data. You should
have an
11
experimental design in mind when you collect the
12
data, and the design itself should be driven by a
13
certain hypothesis. So, this is more
of a
14
philosophical comment.
15
DR. KIBBE: Off-line, we will talk
about
16 the
philosophy of making observations about your
17
surroundings and then developing the thesis and
18
hypothesis versus having an hypothesis and making
19
your observations fit it.
20
Shall we go on to the next question?
21
DR. COONEY: I can't let this
point go
22
unnoticed. I think that you are absolutely right,
23 and
I like your hypothesis, and I think the
24
experiment that has to be done to confirm it is to
25
increase the salaries of professors.
291
1
[Laughter.]
2
DR. KIBBE: What processes in biological
3
manufacturing would benefit the most from
4
implementation of PAT? I think we
are dealing with
5
fermentation here, and the alcohol is just
6
naturally connected somehow. Go
ahead.
7
DR. COONEY: I would address this
question
8 in
two ways. One is if you look at which
9
particular products might benefit by early
10
implementation, and I would suggest that the
11
simpler the better, better to walk rather than run,
12
taking very complex biological products made by
13
very complex processes would be a very perhaps
14
difficult place to begin, so that I think one needs
15 to
think about what are the logical targets.
16
But then within the process, it is
17
important to think about it, as well, because it's
18 in
the fermentation that you define the initial
19
product that is being made, but a lot of the
20
concerns about process variance occur once you have
21
made the product and it is then subsequently being
22
processed.
23
So, that suggests that you need to
24
methodically think through your entire process as
25 you
do for a drug substance, but because there are
292
1
more steps, there is more complexity, obviously,
2
there is more to do.
3
But one of the characteristics of
4
biologicals is that it is important to get the
5
synthesis right, and then it is important to treat
6 it
right once it has been made.
7
DR. DeLUCA: Let me just add to
that, I
8
think it's a good follow-on, and I guess we have
9
heard a lot about the biological process in
10
fermentation, purification, and certainly we have a
11
sensor technology in the Smart systems today to be
12
able to handle that.
13
I guess I wanted to move to the fill and
14
finish end of it, and I guess I have a lot of
15
questions. You know, do you apply
PAT to current
16
products? What properties can
vary from unit to
17
unit? What does the variation
mean in the
18
pharmacological sense? These are
the types of
19
questions.
20
But with regards to the biological, it
21
seems in the fill and finish that most of these
22
freeze dried, so they are going to be lyophilized,
23 and
I think in lyophilization, this is not a
24
trivial situation here, and I think you think that
25 you
put 10,000 vials into a chamber and you get out
293
1
10,000 vials with little variation.
2
You have to look at processing, and most
3
products are processed and then filled in the
4
containers. In freeze drying, the
processing takes
5
place in an individual container.
6
Each little container is processed after
7
it's filled, and the heat that goes to that
8
container is such that each vial doesn't see the
9
same temperature, you like to have a small
10
variation across the shelves in the type of flow,
11 and
the heating element, the fluid that goes
12
through it to heat and freeze, but it isn't.
13
So, you end up with, you could have
14 product that has a variation in moisture, you
can
15
have products that vary in meltback collapse, so
16
this can occur, so I think it is important that
17
when we are looking at PAT and looking at the fill
18 and
finish, that moisture becomes very, very
19
important in these lyophilized products and these
20
biologicals, and I think applying things like NIR
21 to
that, I think makes this a very doable thing, to
22 be
able to do that with every product. It's a
23
non-invasive procedure, and I think that is
24
critical.
25
So, you have to bring in robustness in
294
1
here. What can the product
chemically tolerate in
2 the
way of moisture? I mean it is being freeze
3
dried because obviously, it can't be put into a
4
solution form, so moisture is going to have an
5
effect.
6
But what moisture content, can it tolerate
7 5
percent or maybe it only can tolerate a half
8
percent, or maybe there is an optimum moisture
9
content that is good, because you are going to get
10
into changes in tertiary and quaternary structure,
11
aggregation, and whatnot with regards to moisture.
12 So, I think that is an area that really
13
lends itself to PAT, I think is in actually
14
determining the moisture of these products, and
15
again knowing where, you know, what kind of
16
variation it can tolerate, and you have to somehow
17 try
to have some idea of the pharmacological effect
18 of
this, whatever the effect is of the moisture,
19
does it really translate into a pharmacological
20
effect, but I think that is an area that needs to
21 be
looked at.
22
DR. KOCH: I would certainly agree
with
23
that, but one other part of the biological drug
24
manufacture, particularly fermentation, that I
25
think needs to be addressed is just the
295
1
fermentation itself in terms of reaction
2
engineering.
3
If you look at today's fermenters, they
4
don't look that much different than they did 50, 60
5
years ago. Aeration is very
important, as it
6
nutrient and contact, so maybe the most effective
7
fermentation is where you optimize those
8
parameters, and huge vessels are not necessarily
9 the
way to do that.
10
I think we are in a sunk capital situation
11
where industry probably can't afford to redesign
12 the
approach, but there are some very interesting
13
approaches, that if you could number up from the
14
micro scale that Charles indicated, you might have
15 a
far more effective control of the material and
16 far
less of impurities that are being generated.
17
DR. COONEY: I would like to add
another
18
point to what Pat said. When one
is doing a
19
de-bottlenecking exercise on a manufacturing
20
process to try and improve the throughput, you
21
begin that exercise from the end of the process and
22 you
work your way from the end forward.
23
One very logical way of thinking about the
24
application of the strategy, of PAT strategy, it is
25 due
to exactly the same thing, that if you
296
1
understand, if you really understand the product
2 and
then you work your way back down the process,
3
that makes it easier to de-bottleneck and design
4
going forward.
5
DR. KIBBE: Since we started at
the front,
6
then, moving towards the back with the questions,
7 we
will keep going towards the back.
8
The next question is for processes or
9
products that do not currently allow direct product
10
quality monitoring, what other strategies do you
11
recommend for product quality control in addition
12 to
control of in-process parameters?
13
DR. SINGPURWALLA: I like this question.
14
DR. KIBBE: He likes this
question. It's
15 a
Bayesian question.
16
DR. SINGPURWALLA: Exactly. I am really
17
impressed with your insight and intuition.
18
There is a technology, and there is a
19
technology called information fusion.
Sometimes it
20 is
called information integration. The
basic idea
21 is
this. The analogy here is like
investigating a
22
crime. Some crime has been
committed. You don't
23
know who has committed the crime. You are gathering
24 all
kinds of evidence, and then you are pooling
25
that evidence in a very systematic way to make a
297
1 probabilistic judgment about the crime. You cannot
2
make a judgment with certainty, because the only
3 way
to make a judgment to certainty is to see
4
something.
5
So, you have a similar situation here, and
6 the
problem you mention is common in other
7
scenarios where you cannot directly observe the
8
product.
9
You are not allowed to either observe the
10
product or test the product for whatever reason you
11
have, but you have evidential information. You
12
have information on degradation, you have
13
information on other kind of attributes, and how do
14 you
systematically integrate that information is a
15
well-developed technology, and I think that would
16 be
germane here to the kind of question that you
17 are
raising.
18
You have a process, you cannot directly
19
observe it, but you presumably can observe other
20
things related to it. So, the
question is how do
21 you
systematically pool that information, and there
22 is
a methodology, and, of course, it is Bayesian,
23 as
our chairman so wisely suggested, and it is
24
available.
25
DR. KIBBE: Do you have a
different
298
1
method?
2
DR. LAYLOFF: No, I like Bayesian,
but I
3 was
thinking that the monitoring and control, the
4
in-process parameters may be polyvariate with
5
respect to the quality of product.
6
It may be a series of interactions on
7
product quality, so the thing has to be linked
8
together, so the evidentiary procedure may be many,
9
many different mixtures of it to relate to the
10
product quality.
11
DR. SINGPURWALLA: Just to add to
that,
12 you
used the word polyvariate?
13
DR. LAYLOFF: No, I didn't.
14
[Laughter.]
15
DR. SINGPURWALLA: It's
multivariate.
16
DR. LAYLOFF: I know I did
something
17
wrong.
18
DR. SINGPURWALLA: The thing you
want to
19 be
careful about is this multivariate information
20 may
be interdependent because the same phenomena
21 can
appear under two guises, so you don't want to
22 add
up, you know, the basic information should not
23 be
added up. You have got to recognize the
24
interdependence when there is a multivariate case,
25 and
therefore the technology, the mathematical
299
1
technology that you need has to be nicely refined
2 and
carefully thought out, but the technology is
3
what I am suggesting is purely an analytical
4
technology, it is not a physical technology.
5
DR. WEBBER: Thank you.
6
This question was mostly--I think that's a
7
great answer, but we are looking at sort of, as Dr.
8
Cooney pointed out, the oxygenation issues with
9
fermentation, how that affects product.
If you can
10
understand those, what sort of surrogates can you
11 use
to monitor or have a comfort level with the
12
product quality based on looking at secondary
13
parameters, but I think the answer you have given
14 is
one that we have to consider, as well, the
15
analytical methods used to ensure that those aren't
16
interfering with one another.
17
DR. COONEY: There is a
fundamental
18
problem with surrogates, and that is that many of
19
them come from correlative observation, and the
20
point was made quite appropriately earlier that
21
when you take the data, develop a correlation, that
22 may
work within a certain amount of space with a
23
certain set of assumptions, but it indeed is a
24
correlation.
25
I think the challenge that we have is to
300
1
take that correlative knowledge and then create an
2
hypothesis that, in fact, can be tested by one or
3
more of the many techniques that we are talking
4
about.
5
I think what we are really talking about
6 in
this initiative is a change in the mind-set and
7 the
way that we think about developing and
8
exploring and validating our processes, so there is
9
going to be a lot of these iterations of learning,
10
many of which will come from the surrogate
11
procedures and correlative observations, but we
12
need to drill down and understand that we are
13
solving the right problem at the right time in the
14
right way.
15
DR. KIBBE: Ajaz.
16
DR. HUSSAIN: I think the points
are well
17
made. As we were putting the
guidance together,
18 one
of the key aspects that we did say that in some
19
cases, correlations would not be sufficient from a
20
regulatory perspective, and we would look for
21
causality as a means for making judgments, and so
22
forth, and we sort of leaned that way, risk in a
23
systems way, because keeping in mind, you are
24
looking at a constrained space once it's in
25
manufacturing.
301
1
So, there are opportunities to utilize
2
correlations in some low-risk areas, but also when
3
there is a risk associated, you might prefer it to
4 be
causal rather than correlative. That is
how we
5
sort of structured the guidance.
6
DR. KOCH: I guess this supports
the
7
Bayesian approach, as well, we will find out in a
8
second, but rather than the crime analogy, I would
9
like to think of something that is more like going
10 to
have a physical with a physician, where you are
11
actually trying, in our product quality
12
specifications, you know, assuming that is the
13
perfect health you are looking for, but then
14
develop a number of tests that would be analogous
15 to
doing body fluids or x-rays or a bunch of
16
technologies, and then looking at the results that
17 are
coming back like in the physician's office, all
18 of
the tests are not going to be judged equally,
19 but
if you have a blood pressure and a lipid and an
20
EKG, that is out of a predefined specification, you
21
will start to spend your time at that first in
22
order to see how the process is working.
23
So, you will have a lot of data to work
24
from, but you will have to make some assumptions
25
early on in terms of what type of data relate more
302
1
quickly to the final health of the product in this
2
case.
3
DR. KIBBE: Jurgen, do you have
any
4 opinions
so far?
5
DR. VENITZ: To respond I think to
the top
6
four or five questions in terms of how do you link
7 all
of this to the in-vitro potency, I think
8
Patrick alluded to that a little bit when he talked
9
about the fill stage, the late stage.
You are
10
measuring attributes. You may or
may not know what
11
they actually are other than they depend on some of
12
your process variables.
13
My question, as a pharmacologist, is
14 always
so, why should I care about that. It is
15
driven by your ability to measure, not necessarily
16 by
your ability to understand the consequences.
17 You
do know it is affected by your process, but you
18
don't know whether it has any pharmacological
19
consequences that I should care about.
20
So, whenever you are looking at those
21
steps, there has to be a linkage between whatever
22
attributes you have to the ultimate pharmacological
23
activity of the product.
24
So, maybe that is my comment.
25
DR. KIBBE: Thank you, Jurgen.
303
1
DR. WEBBER: I certainly agree
with that
2
completely. I think one can
measure all sorts of
3
things, but you have to look at what the critical
4
product characteristics are, and that is something
5
that has to be determined during the clinical
6
development stage and product development stages.
7
DR. VENITZ: It goes beyond
that. I mean
8
when I listen to the gentleman talk about what
9
process analytical technology does, the way I
10
understand, it is basically a statistical way of
11
relating process variables and their impact, their
12
criticality in terms of other attributes that you
13
measure using some of those sensors that we heard
14
about, but that doesn't tell me whether I should
15
care about any of this, because you are really then
16
changing your variables to affect your attributes
17 in
a way that you think it should be.
18
But my question is, so what is your
19
template, how do you know that that is the way your
20
attributes should be, unless it is relevant to the
21
pharmacologic activity of your product?
22
If it is not relevant, then, yes, you
23
might be improving your process, but it is
24
cosmetic, it is not of any particular relevance for
25 me
to care about. So, as part of this in
the
304
1
development stage, maybe not in manufacturing, the
2
ultimate manufacturing stage, there has to be a
3
linkage to the pharmacologic activity.
4 DR. KIBBE: Go ahead, Pat.
5
DR. DeLUCA: Just to add, I think
it is
6
critical that we have good control, I mean with
7
regards to the fermentation before it gets to the
8
fill and finish. I think that is
essential. I
9
think when the product gets to the point where it
10 is
going to be formulated and put into final dosage
11
form, you want to have good control over that,
12
because in that finishing, there could be a lot of
13
variation.
14
I just mentioned moisture before, but when
15 you
are freezing, you are freezing an amorphous
16
form, and then when you are drying, you get a lot
17 of
conversion into the crystalline states, and
18
there could be variation in the distribution of
19
crystalline and amorphous form in the finished
20
product, and that is going to affect the water
21
content and the effects, so there is a lot of
22
variation that can occur in that stage, that I
23
think there needs to be control over.
24
I support that there needs to be great
25
control in the upstream processing, as you
305
1
mentioned.
2
DR. KIBBE: Anybody else on the
first
3
four?
4
That brings us to: What
additional
5
elements should be incorporated in a training and
6
certification program for reviewers and inspectors
7 of
biotechnology PAT?
8
I am going to take the prerogative of the
9
Chair to speak first, and then you all will tell me
10
what mistakes I have made.
11
I have been involved with educating
12
college level students for years, and one of the
13
things that I find is that they learn something in
14 a
specific incidence or example, but they don't
15
learn things in the generalities, and one of the
16
most powerful tools we have to handle all of this
17 is
a true and real understanding of the scientific
18
method and the application of the scientific method
19 to
the problem in front of you, and not just
20
learning how to do it in one situation, but
21
learning how that works in any situation.
22
I don't know whether you can get that into
23 a
training session, with a process that hasn't been
24
well developed and isn't templated, when you look
25 at
the data you have in front of you and can use
306
1
that.
2
The second is critical thinking.
Often,
3
students really don't know the difference between
4
facts and opinions, because they have been trained
5
over their life to accept someone who stands in
6 front
of them in a lecture hall and says something
7 as
if they were saying facts, when, in fact, most
8 of
the time we who lecture give opinions about
9
everything and very little real facts.
10
They also have a hard time differentiating
11
results from conclusions. They
look at the result
12 and
immediately leap to a conclusion that isn't
13
necessarily supported by the facts, and if we could
14 add
anything to anybody's training who are going to
15 be
involved in reviewing data or doing inspections,
16 it
is that level of sophistication that would be
17
very helpful.
18
DR. SINGPURWALLA: I agree with
you on
19
that point certainly. I really
also think that
20
there should be some kind of education and training
21 on
basically uncertainty, what is the meaning of
22
uncertainty, how to quantify uncertainty, what are
23 the
different ways of quantifying uncertainty, what
24 is
the difference between variability and
25
uncertainty, is there a difference, and basically,
307
1 not
statistics or statistical technology, but just
2 the
background of what it is all about, I think
3
that would be very valuable because that seems to
4 be
running through completely all the way here.
5
So, that is what I would like to add,
6
"parochial."
7
DR. LAYLOFF: I think also it
would be
8
useful to turn to the biotechnology industry, to
9 the
people working in CMC, to try and help define
10
what attributes reviewers and inspectors should
11
have to properly evaluate, because those guys live
12
with that stuff on a daily basis, and I am sure
13
they would be willing to help.
14
DR. KIBBE: Judy.
15
DR. BOEHLERT: Finally, I am going
to make
16 a
comment. My background is in small
molecules, so
17 I
am sorry, but it would seem to me with the
18
complexity of these processes, that you might want
19 to
go to industry and sort of talk with them, and I
20
think you know, as well, about what are the issues
21
that can occur, because things go wrong in these
22
processes that don't go wrong in conventional
23
processes.
24
You have adventitious contamination and
25
things that don't happen elsewhere, so
308
1
investigators, reviewers need to learn to ask the
2
right questions and go beyond what they see to say,
3
well, what about this, what about that, could this
4
happen here, did this happen here, and that you
5
need to train people to ask the right questions,
6
because it's a whole different ball game when you
7 get
into these products.
8
DR. KIBBE: Anybody else? Do we have
9
anybody who hasn't spoken? Would
you like to
10
comment on our discussion?
11
DR. SELASSIE: I think I will pass.
12
DR. KIBBE: What I am going to do
now,
13
unless Keith has something specific he needs us to
14 do,
is I am going to summarize.
15
DR. WEBBER: I didn't have
anything
16
specific for you to do now. Maybe
before you
17
summarize, I would like to thank the committee
18
certainly for getting together and addressing the
19
issues and the questions that we have, giving
20
presentations, and giving us your input on this
21
difficult issue that we have ahead of us.
22
DR. KIBBE: You had five things
that you
23
wanted us to help you with and some of them we can
24
help you with and sometime we will help you with.
25
Starting with the first one, technology
309
1
changes at an ever-increasing rate since the
2
beginning of civilization, each breakthrough in
3
technology is taking shorter and shorter periods of
4
time. if you wait two weeks,
there will be a new
5
technology to measure something.
6
The question is what do you need to know,
7
what good questions have you asked, and that is the
8
core of the quality of scientific endeavor, so make
9 sure
you ask good questions, and there will be
10
someone out there will develop a way of getting you
11 an
answer.
12
No. 2, data collection is important.
I
13
think there is fun in data mining, I enjoy it. I go
14
looking for patterns and try to develop patterns,
15 but
the question really is are these patterns of
16
correlation of cause and effect, how do you know
17 the
cause and effect, and this boils down to being
18
able to think critically about the analytical data
19 in
front of you.
20
No. 3, variability control seems to be the
21
key. If we know how much
variation we can allow in
22 any
critical step in order to still maintain a good
23
product, then, that is the variation we should
24
allow, and we should really look at variability on
25
each critical step in the process.
310
1
If we know our critical steps, which is
2
always an assumption that we make, and we hope we
3 do,
and we know the variability that will throw our
4
process out of control, then, we know where we need
5 to
limit ourselves, and making intelligent choices
6
about those limits are really important.
7
Related observations. I know my
friend
8
likes, under No. 4, Bayesian approach.
It all
9
boils down to critical thinking about the things
10
that you can measure and the things that you need
11 to
measure, just because you can measure it doesn't
12
mean you need to know about it.
If it is something
13
critical you need to measure, you need to find a
14 way
to measure it.
15
This brings us to No. 5, which I think
16
boils down to training people to think critically
17 and
to apply the scientific method appropriately.
18 The
quality of good science is the quality of the
19
questions.
20
The difference between a normal researcher
21 and
Albert Einstein is that the way he posed the
22
questions allowed him to get breakthroughs and
23
answers. The other thing that he
had that most of
24 us
don't have, and I won't say any one of you
25
doesn't, is that he was never satisfied with the
311
1
quality of the answer, and he always kept looking
2 for
better and better answers.
3
Twain said that what we don't know doesn't
4 get
us in trouble. It is what we know that
ain't
5 so.
6
So, we have to be very careful to avoid
7
thinking we know something about our process just
8
because we made a measurement, and it really isn't
9
something that describes the process, but it is
10
just a convenient measurement.
11
We find in clinical realm that often a
12
technique comes along looking for a disease to
13
diagnose, so don't look for a technique that
14
diagnoses a disease you don't have or wouldn't even
15
get. Just look for the ones that
help you get the
16
answers.
17
On that note, I will end whatever
18
soliloquy I have.
19
Helen, do you have a comment? I
saw you
20
getting closer to the microphone.
21 MS. WINKLE: No.
22
DR. KIBBE: We are going to be
back here
23 at
8:30 tomorrow morning.
24
[Whereupon, the meeting was recessed at
25
4:30 p.m., to reconvene at 8:30 a.m., Wednesday,
312
1
April 14, 2004.]
2 - - -